Study of forestry machines classifiers based on logistic regression

The study aims on development of the approach to classification of forestry machines basing on machine learning methods, which use data clustering and logistic-regression classifier. The study uses information on characteristics of 39 models of eight-wheeled forwarders manufactured by Ponsse, Rottne, Komatsu, John Deere, HSM, Ecolog. The data was obtained from the official websites of the manufacturers and is current for 2023. The data processing was performed on Jupyter Notebook platform. In computational experiments, the classifier model based on logistic regression was able to classify forwarders’ models to one of the four classes corresponding to the cluster labels obtained with k-means algorithm, given the information on 14 passport parameters of the machines, with accuracy at level of 92– 94%. The classifier model, which used as the parameters 1–2 of the principal components of the features matrix, was able to classify the considered forwarder models into 3 classes with accuracy of 97–98%. The accuracy of the model when classifying into 4 classes was lower with estimate at level of 90–94%. Meanwhile, results of the computational experiment showed that logistic regression models when classifying the forwarders into more than 5 classes bear significant loss of the accuracy. The computations results stayed stable during cross-validation with a high assessment of the accuracy. The prospect of further research lays in testing the classifiers on a generated data to assess the sensitivity of the logistic models to the parameters’ changes, and development of a method for obtaining an adequate estimate of missing feature values.

1. Andronov A.V., Petrosyan S.S., Egorin A.A., Ilyushenko D.A., Khitrov E.G. Classification of forwarders using clustering of data on their performance. Resources and Technology, 2021, no. 4, vol. 18, pp. 1–16. (In Russ.)

2. Fedoseeva M.A., Kiselev P.S., Naganov A.S., Petrosian S.S., Schschukin A.V., Khitrov E.G. Classification of 8x8 wheeled forwarders based on cluster analysis in principal component space. Izvestia Sankt-Peterburgskoj Lesotehniceskoj Akademii, 2023, iss. 243, pp. 240–252. (In Russ.)

3. Komatsu official website. URL: https://www.komatsuforest.com/forestmachines/our-forwarders (accessed January 03, 2023). (In Russ.)

4. John Deere official website. URL: https://www.deere.com/en/forwarders/(accessed January 03, 2023). (In Russ.)

5. Ponsse official website. URL: https://www.ponsse.com/en/products/forwarders#/(accessed 03/01/2023)

6. Ecolog official site. URL: https://ecologforestry.com/en/products/forwarders/(Accessed: 03/01/2023)

7. Official website of the HSM company. URL: https://www.hsm-forest.net/forwarders.html (Date of access: 03/01/2023)

8. Official website of the company Rottne. URL: https://www.rottne.com/en/skogsmaskiner/skotare/ (Accessed: 03/01/2023)

9. Website Machine Learning in Python. URL: https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression (Accessed: 08/01/2023)

10. Website Machine Learning in Python. URL: https://scikit-learn.org/stable/modulesclustering.html#k-means (Accessed: 08/01/2023)

11. Website Machine Learning in Python. URL: https://scikit-learn.org/stable/modules/linear_model.html#using-cross-validation (Accessed: 08/01/2023)

12. Nordfjell T., Öhman E., Lindroos O., Ager B. The technical development of forwarders in Sweden between 1962 and 2012 and of sales between 1975 and 2017. International Journal of Forest Engineering, 2019, no. 30:1, рр. 1–13.