Category: 3. Business

The primary goal of this study was to predict muscle response to FES therapy from baseline sEMG signals for individuals with cervical SCI. sEMG has the potential to be integrated into point-of-care tools to provide biomarkers for clinical decision support and enable precision rehabilitation approaches. By knowing which muscles are likely to benefit, therapists can make informed decisions into the use of the limited therapy time available. We evaluated classification models trained on clinical variables alone, sEMG features alone, and combinations of both. Leave-one-participant-out (LOPO) cross validation was used to assess robustness across participants.

Our findings indicate that sEMG feature sets, particularly the FWD feature set (slope sign changes, mean and median frequencies, and M2), consistently outperform models using clinical variables alone. The FWD feature set combined with a random forest classifier achieved the highest performance across multiple metrics, including MCC (0.41), accuracy (0.76), and macro F1 (0.68). Combining clinical variables with sEMG features did not improve model performance, underscoring the unique predictive values of sEMG features. Additionally, training models on specific AIS subgroups (motor complete and motor incomplete) improved performance, particularly in AIS A-B, compared to models trained on the entire dataset.

Limitations of clinical variables alone

Although the progression of AIS grade from A to D correlates with an increasing percentage of responders (Fig. 1), AIS was not identified as the single most important predictor in logistic regression with forward feature selection. Resulting clinical variables also include NLI, Proximity, and Distance. The logistic regression model trained on these variables demonstrated moderate performance (Table 2), with MCC, macro F1 score, accuracy, precision, and TNR all above expected values by chance. Recall (0.21) from the model is below the expected chance level (0.33), indicating the difficulty in correctly identify true responders. This low recall suggests that some muscles that could benefit from the FES therapy might be overlooked.

When evaluating the logistic regression model’s performance per AIS grade (Fig. 3), the limitations of relying solely on clinical variables become more evident, particularly for AIS A, B, and C, where MCC, precision, and recall are all zero. The poor performance on AIS B is especially notable, as it includes a larger number of participants compared to AIS A and C. Training the model on AIS subgroups slightly improved performance for the motor incomplete group (AIS C-D) but did not enhance results for motor complete group (AIS A-B). In fact, the model predicted all AIS A-B muscles to be non-responders, resulting in 14 (25%) false negatives. With no positive predictions, MCC, precision, and recall remained at zero. By relying on this model with only clinical variables, no muscles from patients with AIS A or B would be selected for FES therapy—or potentially for other treatment as well.

Prediction with sEMG features

In contrast, models using sEMG features alone consistently outperformed those based on solely clinical variables, particularly the FWD set with RF. The FWD set achieved the highest values in all metrics, including MCC (0.41), accuracy (0.76), and macro F1 (0.68), though it did not achieve the highest TNR due to the class imbalance. When trained on all participants, FWD set was the only one to obtain above-chance performance across all AIS grades (Fig. 2), suggesting that it captures essential sEMG characteristics relevant to predicting the response to the FES therapy.

The FWD feature set (SSC, M2, mean and median frequencies) captures diverse aspects of motor unit firing and recruitment patterns by integrating both time- and frequency-domain information. This breadth is likely key to its strong predictive performance, as no single feature alone can fully characterize the neuromuscular output, especially after SCI. For example, M2, a time-domain feature characterizing frequency-domain behavior, quantifies the temporal variability of the signal by computing the squared difference between consecutive time samples. Higher M2 values may reflect more abrupt signal changes and complex activation patterns, which contributes towards predicting positive (responder) class in the SHAP analysis (Appendix C). SSC is related to the frequency of slope changes and indicative of motor unit firing irregularity. Mean and median frequencies, which summarize the distribution of spectral power, also contributed but with greater variability across samples. These findings highlight the physiological relevance of the selected features and support the use of diverse and broad feature sets to improve prediction robustness in the heterogeneous SCI population.

Moreover, Fig. 3 shows that training models specifically on motor completeness subgroups (AIS A-B vs. C-D) leads to further performance improvement for RF with the FWD set. This subgroup-specific training enhances MCC, accuracy, macro F1, precision, and TNR, particularly for AIS A-B, highlighting the advantages of tailoring models to motor completeness levels. This approach appears to capture more distinct sEMG patterns within each subgroup, allowing for improved classification performance. Given the small dataset, we did not separate subgroups further by individual AIS grade, though this may provide further improvements with a larger sample size.

Impact of imbalanced dataset on recall (TPR) and TNR

The consistently high TNR across feature sets and models can be attributed to the dataset’s class imbalance, where non-responders are more frequent than responders. This imbalance leads to models that are effective in identifying non-responders but struggle with recall, particularly in AIS A and C, where MCC scores were close to or below zero (Fig. 2) for most models. This low recall indicates that while models perform well in identifying non-responders, they may overlook true responders, limiting the practical utility. Results from Fig. 4 suggests that subgroup-specific training partially alleviates this issue by improving recall within more homogeneous groups, especially in the motor incomplete subgroup, where the model appears better suited to capturing true responder characteristics.

Variability across muscles and participants

Precision variability across participants from RF on the FWD set (Fig. 5) underscores the challenge of achieving consistent responder classification. Precision had the highest variability, followed by MCC, TNR and recall. This variability suggests that while the FWD feature set with RF generally performs well, individual differences in muscle response create inconsistencies in predicting true responders. Notably, the high variability in precision and MCC indicates that certain participants’ sEMG signals are easier to classify than others, possibly due to differences in baseline after SCI or other individual-specific factors. The results in Fig. 5 should however be interpreted with caution, considering the low number of muscles per participant that may impact the robustness of the metrics in this portion of the analysis.

We recognize that differences in muscle type and size may influence sEMG signal characteristic and classification outcomes. Although this variability was not stratified in the current analysis, future studies with more data may explore muscle-specific stratification approaches.

In the context of existing literature

There exists intensive literature in predicting functional recovery after SCI [23, 24, 49]. However, to the best of our knowledge, no prior study exists to provide prediction of muscle response to FES therapy, a promising intervention for restoring motor function. Our study is the first attempt to address the gap in the literature and focuses on muscle-level prediction of FES therapy outcomes. While clinical variables such as AIS grade and NLI provide general prognostic information [22,23,24], our findings indicate that baseline sEMG features, particularly the FWD feature set, are more effective for predicting responses to FES therapy. Unlike available clinical information, sEMG captures neuromuscular activation patterns that reflects residual motor connectivity. The FWD feature set combined with a random forest classifier consistently outperformed models using clinical variables alone, suggesting that sEMG features capture unique, functionally relevant information at the muscle level. Notably, combining clinical variables with sEMG features did not enhance model performance, reinforcing the unique predictive value of sEMG alone.

Prior studies have shown that stratifying SCI patients into specific subgroups based on motor completeness or baseline neurological impairment can improve prognostic accuracy, such as the Unbiased Recursive Partitioning regression with Conditional Inference Trees (URP-CTREE) model [50]. The URP-CTREE model has been used to stratify patients with acute traumatic SCI into homogeneous subgroups to optimize recovery predictions and enhance the design of clinical trials. We explored training separate models for motor complete (AIS A-B) and motor incomplete (AIS C-D) groups. Our findings similarly suggest that subgroup-specific training improves classification performance, particularly in identifying responders, by allowing models to capture subgroup-specific sEMG patterns related to motor completeness.

Choice of MMT as an outcome measure

MMT is a practical and commonly used clinical assessment for muscle strength and was used as the primary outcome measure. Because of its simplicity and accessibility in a clinical setting, MMT was administered before each FES therapy session to track the target muscle strength, without the need for additional scheduling. To ensure consistency across sessions and raters (therapists), we implemented standardized protocols from ISNCSCI and GRASSP.

While not feasible for frequent longitudinal data collection, other modalities such as imaging or motor evoked potentials could provide quantitative insights into muscle structure and corticospinal connectivity in response to FES therapy. Combining these tools with baseline sEMG could offer a comprehensive evaluation framework, capturing both functional and structural aspects of the recovery. A multi-modal approach with measurements before and after FES therapy or at multiple timepoints throughout the therapy cycle could help refine responder identifications and provide more accurate evaluation, enabling more robust predictive model development.

Limitations and future directions

In this section, we discuss several limitations that should be considered when interpreting the findings and future research directions.

Expanding dataset diversity and demographics

First, there was only one female participant (less than 6%), which does not reflect the proportion seen in the SCI population [51] and restricts the study’s generalizability across sex. A more balanced sample would provide a clearer understanding of potential differences in response to FES therapy between male and female participants.

The dataset also has a higher number of non-responders (67%) than responders. This imbalance likely contributed to the high true negative rate (TNR) observed across models, as well as the relatively low recall, indicating that the models may be better at identifying non-responders than true responders. While we attempted to compensate for this effect by evaluating models with robust metrics such as MCC and F1 score, future studies with more balanced responder and non-responder groups would help validate these findings and improve the model’s sensitivity to true responders.

Overrepresentations of AIS D injuries (45%) and C3–C4 level of injury (74%) are also observed. AIS D often reflects more treatment options and better prognostics. Individuals with motor complete cases (AIS A or B) often face lower expected recovery potential. Our results show that clinical information alone typically predicts all muscles in AIS A cases to be non-responders, effectively closing the door to FES therapy for this subgroup. This exclusion is problematic, as AIS A patients represent a group in dire need of interventions. These imbalances hinder the generalizability of findings to the broader SCI population, which exhibits great demographic and clinical diversity.

While we obtained promising results by training models on specific AIS subgroups, the relatively small sample size prevented further stratification by individual AIS grades. Future studies with larger datasets may benefit from more granular subgroup analysis to capture subtle differences in muscle response within each AIS grade, potentially enhancing predictive accuracy.

Although our primary goal in this study was to explore generalizable predictive patterns in baseline sEMG across target muscles for FES therapy, a larger dataset would allow for investigation of the impact from muscle anatomical variability, including muscle type (e.g., biarticular vs uniarticular) and size. Along with sex and other person- and muscle-level variables, muscle type and size could be explored as predictive variables.

Beyond binary classification

In this study, binary classification results were used to evaluate model performance. With a larger dataset, future work could move beyond binary prediction to estimate changes in MMT scores directly. Predicting both the magnitude and the timing of MMT improvement could provide clinicians with more detailed guidance for treatment planning and help set realistic expectations for patients. Also, the confidence level of the prediction could be investigated to indicate the likelihood of responding, providing additional decision support to treating therapists beyond the current binary classification.

Integration of multi-channel perspectives

The experiment was designed with a specific clinical point-of-care implementation in mind: take sEMG measurements from a potential target muscle during voluntary contractions based on MMT protocols, and predict its responsiveness to FES therapy in real time or a short amount of time. As such, simplicity and clinical feasibility were the top priorities—a simple setup with bipolar electrodes and no posture restrictions, with only one recording session. Analysis was also done on individual muscles, instead of multiple muscles together.

While this approach aligns with the implementation goals, it limits the depth of electrophysiological insights. Incorporating signals from multiple channels to analyze agonist and antagonist interactions or co-activation patterns could be beneficial. In our experiments, firing of non-target muscles are often observed even though only the target muscle was voluntarily contracted. Compared to using information solely on the target muscle, these patterns could potentially provide more information regarding the systemic effect of SCI, leading to a more robust muscle-specific prediction.

Acharya, A., Steiner, J. F., Walizada, K. M., Ali, S., Zakir, Z. H., Caiserman, A., and Watanabe, T.: Review article: Snow and ice avalanches in high mountain Asia – scientific, local and indigenous knowledge, Nat. Hazards Earth Syst. Sci., 23, 2569–2592, https://doi.org/10.5194/nhess-23-2569-2023, 2023. a

Afendras, G. and Markatou, M.: Optimality of training/test size and resampling effectiveness in cross-validation, Journal of Statistical Planning and Inference, 199, https://doi.org/10.1016/j.jspi.2018.07.005, 2019. a

Ahsan, M., Khan, A., Khan, K. R., Sinha, B. B., and Sharma, A.: Advancements in medical diagnosis and treatment through machine learning: a review, Expert Syst., 41, e13499, https://doi.org/10.1111/exsy.13499, 2024. a

Amin, G., Imtiaz, I., Haroon, E., Saqib, N. U., Shahzad, M. I., and Nazeer, M.: Assessment of machine learning algorithms for land cover classification in a complex mountainous landscape, Journal of Geovisualization and Spatial Analysis, 8, 34, https://doi.org/10.1007/s41651-024-00195-z, 2024. a

Bakkehøi, S., Domaas, U., and Lied, K.: Calculation of snow avalanche runout distance, Ann. Glaciol., 4, 24–29, https://doi.org/10.3189/S0260305500005188, 1983. a

Bhattacharyya, S.: Confidence in predictions from random tree ensembles, in: 2011 IEEE 11th International Conference on Data Mining, iSSN 2374–8486, https://doi.org/10.1109/ICDM.2011.41, 71–80, 2011. a

Bischl, B., Binder, M., Lang, M., Pielok, T., Richter, J., Coors, S., Thomas, J., Ullmann, T., Becker, M., Boulesteix, A.-L., Deng, D., and Lindauer, M.: Hyperparameter optimization: foundations, algorithms, best practices, and open challenges, WIREs Data Mining and Knowledge Discovery, 13, e1484, https://doi.org/10.1002/widm.1484, 2023. a

Breiman, L.: Random forests, Mach. Learn., 45, 5–32, https://doi.org/10.1023/A:1010933404324, 2001. a, b, c

Bühler, Y., Kumar, S., Veitinger, J., Christen, M., Stoffel, A., and Snehmani: Automated identification of potential snow avalanche release areas based on digital elevation models, Nat. Hazards Earth Syst. Sci., 13, 1321–1335, https://doi.org/10.5194/nhess-13-1321-2013, 2013. a, b

Bühler, Y., von Rickenbach, D., Stoffel, A., Margreth, S., Stoffel, L., and Christen, M.: Automated snow avalanche release area delineation – validation of existing algorithms and proposition of a new object-based approach for large-scale hazard indication mapping, Nat. Hazards Earth Syst. Sci., 18, 3235–3251, https://doi.org/10.5194/nhess-18-3235-2018, 2018. a, b

Campbell, C. and Gould, B.: A proposed practical model for zoning with the avalanche terrain exposure scale, in: Proceedings, International Snow Science Workshop 2013, Grenoble, France, https://arc.lib.montana.edu/snow-science/objects/ISSW13_paper_P5-02.pdf (last access: 1 April 2025), 2013. a

Cetinkaya, S. and Kocaman, S.: IMPACT OF LEARNING SET AND SAMPLING FOR SNOW AVALANCHE SUSCEPTIBILITY MAPPING WITH RANDOM FOREST, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLVIII-M-1-2023, 57–64, https://doi.org/10.5194/isprs-archives-XLVIII-M-1-2023-57-2023, 2023. a

Contreras, P., Orellana-Alvear, J., Muñoz, P., Bendix, J., and Célleri, R.: Influence of random forest hyperparameterization on short-term runoff forecasting in an Andean mountain catchment, Atmosphere-Basel, 12, 238, https://doi.org/10.3390/atmos12020238, 2021. a, b

Copernicus: Tree Cover Density 2018 (raster 10 m and 100 m), Europe, 3-yearly, European Environment Agency (EEA) [data set], https://doi.org/10.2909/486f77da-d605-423e-93a9-680760ab6791, 2020. a

D’Amboise, C., Teich, M., Hormes, A., Steger, S., and Berger, F.: Modeling protective forests for gravitational natural hazards and how it relates to risk-based decision support tools, in: Protective Forests as Ecosystem-based Solution for Disaster Risk Reduction (ECO-DRR), IntechOpen, London, https://doi.org/10.5772/intechopen.99510, 2021. a, b

D’Amboise, C. J. L., Neuhauser, M., Teich, M., Huber, A., Kofler, A., Perzl, F., Fromm, R., Kleemayr, K., and Fischer, J.-T.: Flow-Py v1.0: a customizable, open-source simulation tool to estimate runout and intensity of gravitational mass flows, Geosci. Model Dev., 15, 2423–2439, https://doi.org/10.5194/gmd-15-2423-2022, 2022. a, b, c, d, e

Dankan Gowda, V., Pathak, D., Prasad, K. D. V., Srinivas, V., Manu, Y. M., and Sudhakar Reddy, N.: Scalable machine learning frameworks for large-scale multimodal image and speech signal processing, in: 2024 8th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC), ISSN 2768-0673, https://doi.org/10.1109/I-SMAC61858.2024.10714812, 1693–1699, 2024. a

Dutta, S., Arunachalam, A., and Misailovic, S.: To seed or not to seed? An empirical analysis of usage of seeds for testing in machine learning projects, in: 2022 IEEE Conference on Software Testing, Verification and Validation (ICST), IEEE, Valencia, Spain, https://doi.org/10.1109/ICST53961.2022.00026, 151–161, 2022. a

Engeset, R. V., Pfuhl, G., Landrø, M., Mannberg, A., and Hetland, A.: Communicating public avalanche warnings – what works?, Nat. Hazards Earth Syst. Sci., 18, 2537–2559, https://doi.org/10.5194/nhess-18-2537-2018, 2018. a

Gavaldã, J., Moner, I., and Bacardit, M.: Integrating the ATES into the avalanche information in Aran Valley (Central Pyrenees), in: Proceedings, International Snow Science Workshop 2013, Grenoble, France, https://arc.lib.montana.edu/snow-science/objects/ISSW13_paper_P5-01.pdf (1 May 2025), 2013. a

Gong, Y., Liu, G., Xue, Y., Li, R., and Meng, L.: A survey on dataset quality in machine learning, Inform. Software Tech., 162, https://doi.org/10.1016/j.infsof.2023.107268, 2023. a

Harvey, S., Schmudlach, G., Bühler, Y., Dürr, L., Stoffel, A., and Christen, M.: Avalanche terrain maps for backcountry skiing in Switzerland, in: Proceedings, International Snow Science Workshop, Innsbruck, Austria, Innsbruck, Austria, 1625–1631, https://arc.lib.montana.edu/snow-science/objects/ISSW2018_O19.1.pdf (last access: 12 February 2025), 2018. a

Harvey, S., Christen, M., Bühler, Y., Hänni, C., Boos, N., and Bernegger, B.: Refined Swiss avalanche terrain mapping CATV2/ATHV2, in: Proceedings, International Snow Science Workshop, Tromsø, Norway, 2024, Tromsø, Norway, 1637–1644, https://arc.lib.montana.edu/snow-science/item/3363 (last access: 17 April 2025), 2024. a

Hesselbach, C.: Adaptation and Application of an Automated Avalanche Terrain Classification in Austria, Master’s thesis, University of Natural Resources and Life Sciences, Vienna, https://permalink.obvsg.at/bok/AC16964320 (last access: 22 April 2025), 2023. a, b, c

Horn, B. K. P.: Hillshading and the reflectance map, P. IEEE, 69, 14–47, https://doi.org/10.1109/proc.1981.11918, 1981. a

Huber, A., Hesselbach, C., Oesterle, F., Neuhauser, M., Adams, M., Plörer, M., Stephan, L., Toft, H., Sykes, J., Mitterer, C., and Fischer, J.-T.: AutoATES Austria – testing and application of an automated model-chain for avalanche terrain classification in the Austrian Alps, in: Proceedings, International Snow Science Workshop, Bend, OR, USA, 2023, Innsbruck, Austria, 1272–1278, http://arc.lib.montana.edu/snow-science/item/2989 (last access: 16 March 2025), 2023. a, b, c, d, e, f, g, h, i

Huber, A., Saxer, L., Spannring, P., Hesselbach, C., Neuhauser, M., D’Amboise, C., and Teich, M.: Regional-scale avalanche modeling with com4FlowPy – potential and limitations for considering avalanche-forest interaction along the avalanche track, in: Proceedings, International Snow Science Workshop, Tromsø, Norway, 2024, 587–594, https://arc.lib.montana.edu/snow-science/item.php?id=3194 (last access: 10 February 2025), 2024. a, b, c, d

Jaiswal, J. K. and Samikannu, R.: Application of random forest algorithm on feature subset selection and classification and regression, in: World Congress on Computing and Communication Technologies (WCCCT), https://doi.org/10.1109/wccct.2016.25, 2017. a

Japkowicz, N.: Concept-learning in the presence of between-class and within-class imbalances, in: Advances in Artificial Intelligence, edited by: Stroulia, E. and Matwin, S., Springer, Berlin, Heidelberg, https://doi.org/10.1007/3-540-45153-6_7, 67–77, 2001. a

Joseph, V. R.: Optimal ratio for data splitting, Statistical Analysis and Data Mining: The ASA Data Science Journal, 15, 531–538, https://doi.org/10.1002/sam.11583, 2022. a

Körner, H. J.: The energy-line method in the mechanics of avalanches, J. Glaciol., 26, 501–505, https://doi.org/10.3189/S0022143000011023, 1980. a

Larsen, H. T., Hendrikx, J., Slåtten, M. S., and Engeset, R. V.: Developing nationwide avalanche terrain maps for Norway, Nat. Hazards, 103, 2829–2847, https://doi.org/10.1007/s11069-020-04104-7, 2020. a, b, c, d, e, f, g, h

Markov, K.: kalinmarkov95/machine-learning-auto-ates: Implement Random Forest (RF) for automated ATES classification, Zenodo [code], https://doi.org/10.5281/zenodo.15310357, 2025. a, b

Markov, K. and Ivanov, I.: Avalanche Hazard Assessment and Modeling in Pirin National Park, LOPS Foundation, Sofia, Bulgaria, https://lopsbg.com/wp-content/uploads/2021/09/%D0%9A%D0%B0%D0%BB%D0%B8%D0%BD-%D0%9C%D0%B0%D1%80%D0%BA%D0%BE%D0%B2_%D0%98%D0%B2%D0%B0%D0%BD-%D0%98%D0%B2%D0%B0%D0%BD%D0%BE%D0%B2.pdf (last access: 30 March 2025), 2021. a

Markov, K. and Panayotov, M.: Defining of avalanche terrain with ATES modelling for the region of Bansko Ski Resort in Pirin, in: Avalanches in the Bunderitsa valley in Pirin Mountains, Intel Entrans, Sofia, ISBN 978-619-7703-60-3, 2024. a, b

Markov, K., Panayotov, M., Tcherkezova, E., and Teich, M.: Identifying and mapping avalanche terrain using the ATES model for the region around Bansko Ski Resort, Pirin, Forest Science (Nauka za Gorata), 60, 99–122, 2024. a

Markov, K., Panayotov, M., Tcherkezova, E., and Huber, A.: Influence of forest canopy cover on automated Avalanche Terrain Exposure Scale classification in the Pirin Mountains, Bulgaria, Forestry Ideas, 31, 170–187, https://forestry-ideas.info/issues/issues_Download.php?download=557 (last access: 30 October 2025), 2025. a

McClung, D. and Gauer, P.: Maximum frontal speeds, alpha angles and deposit volumes of flowing snow avalanches, Cold Reg. Sci. Technol., 153, 78–85, https://doi.org/10.1016/j.coldregions.2018.04.009, 2018. a

Nasteski, V.: An overview of the supervised machine learning methods, Horizons, 4, 51–62, https://www.researchgate.net/publication/328146111_An_overview_of_the_supervised_machine_learning_methods (last access: 29 October 2025), 2017. a, b

Oesterle, F., Wirbel, A., Fischer, J.-T., Huber, A., and Spannring, P.: avaframe/AvaFrame: 1.11, Zenodo [code], https://doi.org/10.5281/zenodo.14893015, 2025. a

Panayotov, M. and Tsvetanov, N.: Dating of avalanches in Pirin mountains in Bulgaria by tree-ring analysis of Pinus peuce and Pinus heldriechii trees, Dendrochronologia, 85, 126206, https://doi.org/10.1016/j.dendro.2024.126206, 2024. a

Panayotov, M., Menteshev, B., and Mihaylova, S.: Avalanches in Bulgaria, BEFSA, ISBN 978-619-188-515-2, 2021. a, b, c

Panayotov, M., Christen, M., Bebi, P., and Bartelt, P.: Simulation of avalanches in Bunderitsa valley with the RAMMS software, in: Avalanches in the Bunderitsa valley in Pirin Mountains, Intel Entrans, Sofia, ISBN 978-619-7703-60-3, 2024a. a

Panayotov, M., Markov, K., Tsvetanov, N., Huber, A., Hesselbach, C., and Teich, M.: Avalanche hazard mapping using the avalanche terrain exposure scale (ATES) in the high mountain ranges of Bulgaria, in: Proceedings, International Snow Science Workshop 2024, Tromsø, Norway, https://arc.lib.montana.edu/snow-science/objects/ISSW2024_P12.5.pdf (last access: 21 April 2025), 2024b. a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r

Pandey, S., Chand, S., Horkoff, J., Staron, M., Ochodek, M., and Durisic, D.: Design pattern recognition: a study of large language models, Empir. Softw. Eng., 30, https://doi.org/10.1007/s10664-025-10625-1, 2025. a

Parks Canada Agency: Avalanche Terrain Exposure Scale – Avalanche Terrain Exposure Scale, Government of Canada, https://parks.canada.ca/pn-np/bc/glacier/visit/hiver-winter/ski/ates (last access: 29 October 2025), 2017. a

Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E.: Scikit-learn: machine learning in Python, J. Mach. Learn. Res, 12, 2825–2830, 2011. a

Pirin National Park: Pirin National Park – World Heritage Property of Unesco, https://www.pirin.bg/?p=5591#more-5591 (last access: 28 August 2025), 2022. a, b

Prasetiyowati, M. I., Maulidevi, N. U., and Surendro, K.: Determining threshold value on information gain feature selection to increase speed and prediction accuracy of random forest, Journal of Big Data, 8, 84, https://doi.org/10.1186/s40537-021-00472-4, 2021. a

Probst, P., Wright, M., and Boulesteix, A.-L.: Hyperparameters and tuning strategies for random forest, WIREs Data Mining and Knowledge Discovery, 9, e1301, https://doi.org/10.1002/widm.1301, 2019. a, b

Pugliese, R., Regondi, S., and Marini, R.: Machine learning-based approach: global trends, research directions, and regulatory standpoints, Data Science and Management, 4, 19–29, https://doi.org/10.1016/j.dsm.2021.12.002, 2021. a

Python Software Foundation: Python Language Reference, version 3.9.21 [code], https://www.python.org/ (last access: 22 March 2025), 2024. a

Rainio, O., Teuho, J., and Klén, R.: Evaluation metrics and statistical tests for machine learning, Sci. Rep.-UK, 14, 6086, https://doi.org/10.1038/s41598-024-56706-x, 2024. a, b, c

Ramadhan, M., Sitanggang, I., Nasution, F., and Ghifari, A.: Parameter tuning in random forest based on grid search method for gender classification based on voice frequency, DEStech Transactions on Computer Science and Engineering, https://doi.org/10.12783/dtcse/cece2017/14611, 2017. a

Rimal, Y., Sharma, N., and Alsadoon, A.: The accuracy of machine learning models relies on hyperparameter tuning: student result classification using random forest, randomized search, grid search, bayesian, genetic, and optuna algorithms, Multimed. Tools Appl., 83, 74349–74364, https://doi.org/10.1007/s11042-024-18426-2, 2024. a

Rogers, J. and Gunn, S.: Identifying feature relevance using a random forest, in: SLSFS: International Statistical and Optimization Perspectives Workshop “Subspace, Latent Structure and Feature Selection”, edited by: Saunders, C., Grobelnik, M., Gunn, S., and Shawe-Taylor, J., Springer, Berlin, Heidelberg, https://doi.org/10.1007/11752790_12, 173–184, 2006. a

Salman, H. A., Kalakech, A., and Steiti, A.: Random forest algorithm overview, Babylonian Journal of Machine Learning, 2024, 69–79, https://doi.org/10.58496/BJML/2024/007, 2024. a, b

Saraswat, P.: Supervised machine learning algorithm: a review of classification techniques, Smart Innovation, Systems and Technologies, 273, 477–482, https://doi.org/10.1007/978-3-030-92905-3_58, 2022. a

Sarker, I. H.: Machine learning: algorithms, real-world applications and research directions, SN Computer Science, 2, 160, https://doi.org/10.1007/s42979-021-00592-x, 2021. a

Schmudlach, G. and Köhler, J.: Method for an automatized avalanche terrain classification, in: Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 2016, Breckenridge, Colorado, 729–736, https://arc.lib.montana.edu/snow-science/objects/ISSW16_P2.04.pdf (last access: 7 April 2025), 2016. a

Schumacher, J., Toft, H., McLean, J. P., Hauglin, M., Astrup, R., and Breidenbach, J.: The utility of forest attribute maps for automated Avalanche Terrain Exposure Scale (ATES) modelling, Scand. J. Forest Res., 37, 264–275, https://doi.org/10.1080/02827581.2022.2096921, 2022. a, b, c, d, e

Schweizer, J. and Lüschg, M.: Characteristics of human-triggered avalanches, Cold Reg. Sci. Technol., 33, 147–162, https://doi.org/10.1016/S0165-232X(01)00037-4, 2001. a

Schweizer, J., Bartelt, P., and van Herwijnen, A.: Snow avalanches in: Snow and Ice-Related Hazards, Risks and Disasters, Elsevier, https://doi.org/10.1016/b978-0-12-394849-6.00012-3, 395–436, 2015. a

Sharp, A. E. A.: Evaluating the exposure of heliskiing ski guides to avalanche terrain using a fuzzy logic avalanche susceptibility model, Masters’ thesis, University of Leeds, School of Geography, https://doi.org/10.13140/RG.2.2.18673.94567, 2018. a

Sharp, E., Campbell, C., Statham, G., and Schroers, B.: A standardized, multiscale, fuzzy spatialdata model for a Avalanche Terrain Exposure Scale mapping, in: Proceedings, International Snow Science Workshop, Bend, Oregon, 2023, 838–845, http://arc.lib.montana.edu/snow-science/item/2974 (last access: 5 April 2025), 2023. a

Siji George, G. C. and Sumathi, B.: Grid search tuning of hyperparameters in random forest classifier for customer feedback sentiment prediction, International Journal of Advanced Computer Science and Applications, 11, https://doi.org/10.14569/IJACSA.2020.0110920, 2020. a

Spannring, P.: Comparison of two avalanche terrain classification approaches: Avalanche Terrain Exposure Scale – Classified Avalanche Terrain, Master’s thesis, University of Innsbruck, Faculty of Geo- and Atmospheric Sciences, https://resolver.obvsg.at/urn:nbn:at:at-ubi:1-155858 (last access: 30 October 2025), 2024. a

Statham, G. and Campbell, C.: The Avalanche Terrain Exposure Scale (ATES) v.2, Nat. Hazards Earth Syst. Sci., 25, 1113–1137, https://doi.org/10.5194/nhess-25-1113-2025, 2025. a, b, c, d, e, f

Statham, G., McMahon, B., and Tomm, I.: The Avalanche Terrain Exposure Scale, in: Proceedings, International Snow Science Workshop 2006, Telluride, Colorado, https://arc.lib.montana.edu/snow-science/objects/issw-2006-491-497.pdf (last access: 5 May 2025), 2006. a, b, c

Sykes, J., Toft, H., Haegeli, P., and Statham, G.: Automated Avalanche Terrain Exposure Scale (ATES) mapping – local validation and optimization in western Canada, Nat. Hazards Earth Syst. Sci., 24, 947–971, https://doi.org/10.5194/nhess-24-947-2024, 2024. a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q

Techel, F. and Zweifel, B.: Recreational avalanche accidents in Switzerland: trends and patterns with an emphasis on burial, rescue methods and avalanche danger, in: Proceedings International Snow Science Workshop 2013, Grenoble, France, https://arc.lib.montana.edu/snow-science/item.php?id=1844 (last access: 4 April 2025), 2013. a

Techel, F., Jarry, F., Kronthaler, G., Mitterer, S., Nairz, P., Pavšek, M., Valt, M., and Darms, G.: Avalanche fatalities in the European Alps: long-term trends and statistics, Geogr. Helv., 71, 147–159, https://doi.org/10.5194/gh-71-147-2016, 2016. a

Toft, H. B., Sykes, J., Schauer, A., Hendrikx, J., and Hetland, A.: AutoATES v2.0: Automated Avalanche Terrain Exposure Scale mapping, Nat. Hazards Earth Syst. Sci., 24, 1779–1793, https://doi.org/10.5194/nhess-24-1779-2024, 2024. a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y

Tonnel, M., Wirbel, A., Oesterle, F., and Fischer, J.-T.: AvaFrame com1DFA (v1.3): a thickness-integrated computational avalanche module – theory, numerics, and testing, Geosci. Model Dev., 16, 7013–7035, https://doi.org/10.5194/gmd-16-7013-2023, 2023. a

Tsvetanov, N. and Panayotov, M.: Dendrochronological reconstruction of large magnitude avalanches in Bunderitsa valley, in: Avalanches in the Bunderitsa valley in Pirin Mountains, Intel Entrans, Sofia, ISBN 978-619-7703-60-3, 2024. a, b

Ul Hassan, C. A., Khan, M. S., and Shah, M. A.: Comparison of machine learning algorithms in data classification, in: International Conference on Automation and Computing (ICAC), https://doi.org/10.23919/IConAC.2018.8748995, 2018. a

Varoquaux, G. and Colliot, O.: Evaluating machine learning models and their diagnostic value, in: Machine Learning for Brain Disorders, Springer US, ISBN 978-1-0716-3195-9, 2023.  a, b, c

Veitinger, J., Purves, R. S., and Sovilla, B.: Potential slab avalanche release area identification from estimated winter terrain: a multi-scale, fuzzy logic approach, Nat. Hazards Earth Syst. Sci., 16, 2211–2225, https://doi.org/10.5194/nhess-16-2211-2016, 2016a. a, b

Veitinger, J., Sovilla, B., and Purves, R. S.: Model code of release area algorithm [code], https://github.com/jocha81/Avalanche-release (last access: 30 October 2025), 2016b. a, b

Venkadesh, P., S. V., D., Marymariyal, P., and Keerthana, S.: Predicting Natural Disasters With AI and Machine Learning, IGI Global Scientific Publishing, https://doi.org/10.4018/979-8-3693-3362-4.ch003, 2024. a

Viallon-Galinier, L., Hagenmuller, P., and Eckert, N.: Combining modelled snowpack stability with machine learning to predict avalanche activity, The Cryosphere, 17, 2245–2260, https://doi.org/10.5194/tc-17-2245-2023, 2023. a

von Avis, C. D., Sykes, J. M., and Tutt, B.: Development of large scale automated Avalanche Terrain Exposure Scale (ATES) ratings in collaboration with local avalanche experts, in: Proceedings, International Snow Science Workshop 2023, Bend, Oregon, 982–988, http://arc.lib.montana.edu/snow-science/item/2998 (last access: 12 April 2025), 2023. a, b

Wang, S., Zhuang, J., Zheng, J., Fan, H., Kong, J., and Zhan, J.: Application of Bayesian hyperparameter optimized random forest and XGBoost model for landslide susceptibility mapping, Front. Earth Sci., 9, https://doi.org/10.3389/feart.2021.712240, 2021. a

Werners, B.: Aggregation models in mathematical programming, in: Mathematical Models for Decision Support, Vol. 48 of NATO ASI Series, Springer, Berlin, Heidelberg, 295–305, https://doi.org/10.1007/978-3-642-83555-1_19, 1988. a

Yolanda, A., Adnan, A., Ell Goldameir, N., and Rizalde, F.: The comparison of accuracy on classification data with machine learning algorithms (case study: human development index by regency/city in Indonesia 2020), in: AIP Conference Proceedings, https://doi.org/10.1063/5.0118720, 2023. a

By Kosaku Narioka

  Honda Motor cut its annual earnings forecasts after a weak first half, flagging slumping car sales in China and Southeast Asia and a nearly $1 billion drag due to a shortage of chips from Dutch supplier Nexperia. 

  Executive Vice President Noriya Kaihara said the semiconductor crunch had affected production in North America since last Monday. He said the carmaker is working to restore production in the week of Nov. 21, as shipments of Nexperia chips from China appear to be resuming. China's Commerce Ministry said earlier this month that the country would permit exports of Nexperia chips in eligible cases, without specifying the criteria. 

  The Japanese automaker on Friday estimated an operating profit hit of 150.0 billion yen, equivalent to $980 million, from the chip shortage for the year through March. 

  Honda also lowered its car sales forecast, blaming weaker sales in Asia and the chip crunch amid a dispute between the Dutch and Chinese governments over control of the semiconductor maker. 

  Kaihara said that demand is weaker in some Southeast Asian nations and that competition is intensifying in countries like Thailand as rival carmakers offer sales incentives and cut auto prices to compete with emerging Chinese players. The company needs to make drastic changes in Asia to address weak sales, he said. 

  Honda now expects group car sales of 3.34 million units this fiscal year, down from 3.62 million forecast previously. First-half sales dropped 5.6% to 1.68 million vehicles. 

  Tariffs remained a drag on results, with U.S. duties reducing operating profit by Y164.3 billion for the six months ended September, the company said. However, it projected a smaller tariff burden of Y385.0 billion for the fiscal year versus a previous estimate of Y450.0 billion. 

  Honda's stock has lagged behind the broader market as U.S. tariffs clouded its earnings outlook. Its shares are up about 3% this year compared with the benchmark Nikkei Stock Average's roughly 26% gain. 

  The automaker said Friday that first-half net profit fell 37% from a year earlier to Y311.83 billion. That missed the Y342.97 billion estimate of analysts in a poll by data provider Quick. Revenue declined 1.5% to Y10.633 trillion. 

  Its motorcycle business fared better, with operating profit increasing 13% to Y368.2 billion as higher sales in Brazil and the Philippines offset a decline in Vietnam. 

  The company also booked Y223.7 billion in one-time electric vehicle-related expenses as it provided for losses and impairment on EVs sold in the U.S. and wrote down EV development assets due to lineup changes. 

  Honda said in May that it planned to cut its EV investment by some $20 billion in the coming years as the demand growth slows. The automaker said it would improve its lineup of hybrid models. That came as some consumers in the U.S. and other markets have shifted to hybrids from pure EVs amid concerns about charging problems and higher prices associated with fully electric cars. 

  For the year ending March, the company projected revenue to decline 4.6% to Y20.700 trillion and net profit to drop 64% to Y300.00 billion. It previously projected revenue of Y21.100 trillion and net profit of Y420.00 billion. 

  Honda was the last of Japan's biggest automakers to report earnings. Earlier this week, Toyota Motor posted stronger second-quarter net profit and raised its full-year sales and earnings guidance despite an expected $9 billion blow from U.S. tariffs. On Thursday, Nissan Motor booked its fifth straight quarterly net loss, driven in part by a tariff hit of more than half a billion dollars. 

  Write to Kosaku Narioka at kosaku.narioka@wsj.com 

  (END) Dow Jones Newswires

  November 07, 2025 08:24 ET (13:24 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.

Pelican Bay Capital Management, an investment management company, released its third-quarter 2025 investor letter. A copy of the same can be downloaded here. PBCM Concentrated Value Strategy returned 7.8% in the quarter, compared to a 5.3% return for the Russell 1000 Value Index. YTD, the fund returned 11.2% compared to 11.6% for the index. In addition, please check the fund’s top five holdings to know its best picks in 2025.

In its third-quarter 2025 investor letter, PBCM Concentrated Value Strategy highlighted stocks such as Alphabet Inc. (NASDAQ:GOOG). Alphabet Inc. (NASDAQ:GOOG), the parent company of Google, offers various platforms and services operating through Google Services, Google Cloud, and Other Bets segments. The one-month return of Alphabet Inc. (NASDAQ:GOOG) was 20.15%, and its shares gained 58.65% of their value over the last 52 weeks. On November 6, 2025, Alphabet Inc. (NASDAQ:GOOG) stock closed at $285.34 per share, with a market capitalization of $3.439 trillion.

PBCM Concentrated Value Strategy stated the following regarding Alphabet Inc. (NASDAQ:GOOG) in its third quarter 2025 investor letter:

“Alphabet Inc. (NASDAQ:GOOG) gained 41% this quarter as they also benefited from increasing demand for their AI services. GOOG’s Gemini AI app has recently surpassed OpenAI’s ChatGPT app in the Apple app store, and the company’s Tensor Processing Chips have become a viable alternative to Nvidia’s GPUs in Data Center’s dedicated to AI use. I would note that GOOG’s stock price has increased to the top end of our estimated intrinsic valuation range, and we have trimmed our position meaningfully.”

Alphabet Inc. (NASDAQ:GOOG) is in the 7th position on our list of 30 Most Popular Stocks Among Hedge Funds. As per our database, 178 hedge fund portfolios held Alphabet Inc. (NASDAQ:GOOG) at the end of the second quarter which was 164 in the previous quarter. In the third quarter of 2025, Alphabet Inc. (NASDAQ: GOOG) achieved its first-ever $100 billion in revenue. While we acknowledge the potential of Alphabet Inc. (NASDAQ:GOOG) as an investment, we believe certain AI stocks offer greater upside potential and carry less downside risk. If you’re looking for an extremely undervalued AI stock that also stands to benefit significantly from Trump-era tariffs and the onshoring trend, see our free report on the best short-term AI stock.

In another article, we covered Alphabet Inc. (NASDAQ:GOOG) and shared the list of stocks Jim Cramer discussed. In addition, please check out our hedge fund investor letters Q3 2025 page for more investor letters from hedge funds and other leading investors.

The vital flow of chips from China to the car industry in Europe looks poised to resume as part of the deal struck last week between Donald Trump and his Chinese counterpart, Xi Jinping.

The Netherlands has signalled that its standoff with Beijing is close to a resolution amid signs China’s ban on exports of the key car industry components is easing.

The dispute began when the Dutch government took control of chipmaker Nexperia at the end of September amid US security concerns about its Chinese owner, Wingtech. Beijing retaliated by halting all exports from Nexperia’s factories in the country, threatening to disrupt car production in Europe and Japan.

The White House had put Wingtech on a list of companies that would have their exports to the US controlled under its “affiliate rule”. However, as part of the deal between Trump and Xi in Korea, the US authorities will now delay the implementation of this rule for a year in exchange for China pausing its own restrictions on exports of chips and crucial rare-earth minerals.

The Netherlands’ economy minister, Vincent Karremans, said on Thursday he trusted that Nexperia chips would reach customers in Europe and the rest of the world in the coming days.

Meanwhile, one of the main suppliers to the German car industry, Aumovio, confirmed on Friday it had received notice from China that chips supply would resume to its operations.

“We applied for and received an exemption from the export restrictions. We received it ‌the day before yesterday verbally, yesterday in writing,” the Aumovio chief executive, Philipp von Hirschheydt, said after the company reported ⁠its third-quarter results.

At the heart of the dispute is control of Nexperia’s operations in Nijmegen, the Netherlands, after the company was bought by China’s Wingtech in 2018. Karremans took control of the chipmaker on 30 September, amid fears its operations and intellectual property would be moved to China.

Nexperia in the Netherlands said it was “pleased by the one-year suspension by US authorities of the so-called affiliate rule” and also welcomed “China’s commitment to facilitate the resumption of exports from Nexperia’s Chinese facility”.

But it added there continue to be some concerns and it could tell “when products from our facility in China will be delivered”.

The row that threatened to halt car assembly lines in Europe underlines the global nature of car industry’s supply chain and the vulnerability of European and Japanese companies that rely on China for chips.

US authorities had also raised security concerns about Wingtech and Nexperia’s Chinese chief executive, Zhang Xuezheng, in June, court documents show.

Four days after the seizure, China banned exports from Nexperia’s factories in the country, where about 70% of its chips are packaged before distribution. By the end of last month, Nexperia had retaliated by halting chip supplies to a Chinese plant.

Bloomberg reports on Friday cited sources saying the Dutch government was ready to shelve the order that gave it power to block or change key corporate decisions at Nexperia on the condition that China resumes exports of critical chips.

Karremans said the Netherlands had been informed by China and the US that the deal they struck in Korea last month would enable the resumption of supplies from Nexperia’s facilities in China.

“This is also consistent with information provided to the European Commission by the Chinese Ministry of Commerce,” he said.

By Jamie Chisholm

The No. 1 crypto is closely correlated with the Nasdaq

Liquidity pressures should ease, helping bitcoin rally

The latest stock market pullback has been led by technology plays, it’s pretty clear. The tech-heavy Nasdaq Composite COMP closed Thursday down 3.8% from its record high registered last week, while the broader S&P 500 SPX has retreated 2.5% from its peak.

Why Big Tech has been struggling of late, however, is more open to debate. The most popular theory is that rich valuations can’t cope with burgeoning doubts about returns on AI-linked capital investment.

But strategists at Citi, led by Dirk Willer, are skeptical that recent volatility is because of angst over Big Tech ROI. “We will not wade into this debate, as our best guess is that the market will give the companies some more time before expecting a return on investment,” the Citi team said in a note published late Thursday.

Citi does accept that hyperscalers raising debt on and off their balance sheet – rather than using cash – to pay for the AI build-out is a source of worry. But they argue the main cause of the stock market’s latest wobble is declining financial-system liquidity.

And one of the best ways to track that, they reckon, is via the performance of bitcoin (BTCUSD). The crypto asset this week fell into bear market territory, having lost more than 20% from its recent record high. The move came as the Treasury is rebuilding its general account (TGA), which in effect takes funds from the market. Since mid-July, bank reserves have fallen by around $500 billion, and such a trend has historically impacted bitcoin, according to Citi.

“Traditionally, falling reserves have also impacted equities negatively, but this did not happen prior to this week. But it is plausible that bitcoin is a more sensitive instrument for pure liquidity, especially with equities caught up in the fundamentally-driven AI narrative,” Citi says.

And the problem for tech stocks is that bitcoin acts as a warning signal for the Nasdaq NDX, Citi suggests. “We had shown in the past that NDX trades much better when bitcoin is trading well, and vice versa,” Citi says. “In particular, being long NDX only when bitcoin is above its 55-day moving average (and lagging it by a day) improves the active information ratio for NDX from 0.95 to 1.4 and is similarly significant for longer 2 and 3 day lags.”

The information ratio measures portfolio returns and indicates a portfolio manager’s ability to generate excess returns relative to a given benchmark.

Bitcoin is currently below its 55-DMA. The good news for the crypto, and by extension tech stocks, is that Citi says the TGA has now reached more than $900 billion, a level at which the Treasury typically has stopped the rebuilding process in the post-COVID period.

“This would suggest that liquidity conditions should improve going forward, which should support bitcoin, and could also get the NDX Santa rally back on track,” says Citi.

The markets

U.S. stock-index futures (ES00) (YM00) (NQ00) are lower as benchmark Treasury yields BX:TMUBMUSD10Y rise. The dollar index DXY is up, while oil prices (CL.1) gain ground and gold futures (GC00) are trading around $4,015 an ounce.

   Key asset performance                                                Last       5d      1m      YTD      1y 
   S&P 500                                                              6720.32    -1.75%  2.56%   14.26%   12.09% 
   Nasdaq Composite                                                     23,053.99  -2.24%  0.13%   19.38%   19.64% 
   10-year Treasury                                                     4.11       3.10    7.40    -46.60   -20.00 
   Gold                                                                 4012.5     -0.02%  -0.57%  52.03%   49.07% 
   Oil                                                                  60.16      -1.18%  3.30%   -16.29%  -14.58% 
   Data: MarketWatch. Treasury yields change expressed in basis points 

Need to Know starts early and is updated until the opening bell, but sign up here to get it delivered once to your email box. The emailed version will be sent out at about 7:30 a.m. Eastern.

Take control of your news. Make MarketWatch your preferred source on Google.

The buzz

U.S. economic data due Friday include the University of Michigan consumer sentiment survey for November, released at 10 a.m. Eastern.

Federal Reserve officials speaking Friday include Fed Vice Chair Philip Jefferson at 7 a.m., and Fed governor Stephen Miran at 3 p.m.

Tesla stock (TSLA) is slightly lower after investors approved Elon Musk’s $1 trillion pay package. Separately, Musk said Tesla plans an AI chip fabrication plant in conjunction with Intel (INTC).

The U.S. will block sales to China of some scaled-down Nvidia (NVDA) chips, according to a report.

Peloton Interactive shares (PTON) are jumping in premarket action after the fitness company beat first-quarter fiscal 2026 earnings estimates.

DraftKings stock (DKNG) is falling after the betting group trimmed its full-year sales outlook, as it invests more in prediction markets.

Expedia (EXPE) shares are jumping after the travel group gave upbeat guidance.

Best of the web

Beware the three Ls: leverage, liquidity and lunacy.

Blackstone is offloading a flopped $1.8 billion investment in senior housing.

The man who shaped the internet won’t be able to fix it.

How the lowly soybean got trapped in the crossfire of the U.S.-China trade wars.

The chart

It’s time to buy cyclical stocks says Jim Paulsen. Writing in his Paulsen Perspectives blog, the Wall Street veteran strategist argues that this weeks news of a jump in the Challenger Job Cuts Announcements index makes it more likely that the Federal Reserve will continue cutting interest rates, “helping to spike the punch bowl for cyclical companies.”

“Cyclical stocks have greatly underperformed this year, but with job losses mounting, during the months ahead, even the Mag7 may not be able to keep pace with old-line CYCLICALS!” says Paulsen.

Top tickers

Here were the most active stock-market tickers on MarketWatch as of 6 a.m. Eastern.

   Ticker  Security name 
   TSLA    Tesla 
   NVDA    Nvidia 
   PLTR    Palantir Technologies 
   AMD     Advanced Micro Devices 
   GME     GameStop 
   BYND    Beyond Meat 
   TSM     Taiwan Semiconductor Manufacturing 
   META    Meta Platforms 
   IREN    IREN 
   OPEN    Opendorr Technologies 

Random reads

Did you hear that?! Elf movie costume up for auction.

U.K. housing website’s AI travails highlights adoption angst.

‘Lion’ on the loose in Ireland was a big dog ‘with a fresh haircut.’

For more market updates plus actionable trade ideas for stocks, options and crypto, subscribe to MarketDiem by Investor’s Business Daily.

-Jamie Chisholm

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

  (END) Dow Jones Newswires

  11-07-25 0743ET
Copyright (c) 2025 Dow Jones & Company, Inc.

Under the Electronic Communications Code 2017 (the Code), telecoms operators can ask a property tribunal to impose a so-called ‘Code agreement’ on landowners in the event they cannot agree on such an agreement between themselves. However, where landowners can demonstrate their intent to redevelop all or part of the land to which the desired Code rights would relate, or any neighbouring land, and that they could not reasonably do so if a Code agreement was imposed, the tribunal is prohibited from imposing such an agreement on the parties. This ‘redevelopment defence’ is provided for under Paragraph 21(5) of the Code.

In a recent case ruled on by the First-tier Tribunal (Property Chamber) (FTT), Icon Tower Infrastructure Limited (Icon) sought to resist the imposition of a Code agreement on it in respect of freehold land it owns at Queens Oak Farm in Northamptonshire. On Tower UK Limited (On Tower) has maintained a telecoms mast on a site on that land since around 1997. On Tower’s mast is used by the UK’s biggest mobile network operators (MNOs) – EE, Three, Vodafone, and Virgin Media O2 – for hosting electronic communications apparatus.

On Tower previously held a formal lease to operate from the Icon-owned site, but that lease agreement expired in 2016. Since then, On Tower has been operating from the site under a so-called tenancy at will, which is a form of tenancy that is not subject to a formal lease or end date. Under this arrangement, On Tower pays Icon an annual rent and a proportion of the income it receives from the mobile network operators for use of its mast.

On Tower is seeking a Code agreement to enhance its rights to operate on the site. Compared to the preceding legislative regimes, the 2017 Code is weighted more heavily in favour of telecoms operators than landowners in respect of the rights a Code agreement confers on operators to install, inspect and maintain equipment such as masts, cables and other communications apparatus on others’ land.

On Tower previously won a protracted legal battle that ended up in the UK Supreme Court over its rights to seek a Code agreement with AP Wireless, a company in the same group as Icon, in respect of the Queens Oak Farm site. However, when the case was remitted to FTT, Icon, which was by then the owner of the Queens Oak Farm site, claimed it had a redevelopment defence to defeat the imposition of the Code agreement sought.

As well as being a landowner, Icon is also a telecoms company, part of the AP Wireless group. It has designs on installing its own mast on the site On Tower occupies at Queens Oak Farm and of encouraging the MNOs that use On Tower’s mast currently to switch to its mast. In the latest proceedings in this long-running dispute, the FTT had to decide whether Icon had a legitimate redevelopment defence it could rely on.

The central question the Tribunal had to determine was whether Icon could demonstrate a “firm and settled intention” to redevelop the site, such that it could not reasonably do so if On Tower remained in occupation of the site.

The Tribunal applied a two-stage test to help it answer this question, involving assessment of subjective and objective factors. In respect of the subjective part of the test, the Tribunal considered whether Icon did genuinely intend to redevelop. With the objective part of the test, it considered whether there was a reasonable prospect of Icon being able to carry out the redevelopment.

The Tribunal also considered whether Icon’s intention was “conditional” – i.e. whether its plans for redevelopment were tied to the purpose of defeating On Tower’s bid for Code rights. It further had to determine whether the works Icon planned amounted to genuine “redevelopment”.

On this last point, the Tribunal held that replacing one mast with another can constitute redevelopment under the Code, but only if the legal tests around intent and reasonable prospects are met.

On Icon’s intent, the Tribunal found that the company’s redevelopment plan was investment-led, based on a business plan assuming all MNOs would migrate to the new mast. However, it found no evidence that Icon had actually engaged with the MNOs, and Icon’s own witnesses accepted there was a real risk the MNOs would not move to the new mast. The Tribunal said that while Icon has “a firm and settled intention to carry out its redevelopment” this plan is “wedded to MNO’s migrating from On Tower”.

In considering the likelihood of redevelopment works going ahead, the Tribunal concluded that Icon had not shown a reasonable prospect of carrying out the redevelopment as planned, because “on the balance of probabilities … the most likely outcome is that the MNOs will not migrate to Icon’s new tower”. It reached this view after considering evidence that pointed to Icon’s lack of relationship with the MNOs, the fact Icon has built other “speculative” towers which remain unoccupied, and the fact the MNOs have been working with On Tower to find an alternative site.

The FTT said: “MNOs have not migrated to any of Icon’s new towers. This litigation will have damaged any future relationship Icon may have had with MNOs.”

On the issue of conditionality, the Tribunal accepted that while Icon’s strategy was partly motivated by a desire to remove On Tower as a competitor, this is “a perfectly legitimate business aim” and not improper. However, case law has established that, for the redevelopment defence to be relied upon, the intention to redevelop must exist independently of whether an operator asserts a claim to Code rights – and the Tribunal in this case considered that Icon’s redevelopment plan was so closely tied to the outcome of the litigation that it lacked that necessary independence.

On the issue of conditionality, the Tribunal accepted that while Icon’s strategy was partly motivated by a desire to remove On Tower as a competitor, this is “a perfectly legitimate business aim” and not improper. However, case law has established that, for the redevelopment defence to be relied upon, the intention to redevelop must exist independently of whether an operator asserts a claim to Code rights – and the Tribunal in this case considered that “Icon would intend to do the same works” even if On Tower did not seek Code rights.

As a result of its findings, the Tribunal held that Icon had not established a genuine and deliverable intention to redevelop within the meaning of paragraph 21(5) of the Code. As such, its redevelopment defence failed. The Tribunal ruled that the statutory test for imposing a new Code agreement in favour of On Tower was met.

Property dispute resolution specialist Mairghread Yule of Pinsent Masons, who acted for On Tower in the case, said: “This decision will be welcomed by Code operators. This judgment will be of wide interest and application in the industry, especially regarding redevelopment. It provides useful findings on redevelopment intention – subjective, objective and conditionality intention – and how this will be assessed and considered by the judiciary.”

Ian Morgan, who was part of the Pinsent Masons team involved in the earlier Supreme Court proceedings, added: “This decision will be of significance not only to parties dealing with the Electronic Communications Code, but also because it considers in some detail significant case law relevant to the Landlord and Tenant Act 1954, which may be of broader appeal.”