Mid Sweden University

This report aims to solve a problem for the   waxers in the Swedish National Cross-country Ski Team, which hereafter will   be referred to as the national team. The problem in hand is that currently,   the national team lacks a system for book-keeping of ski pairs and ski tests.   Also, the project intends to provide a tool for predicting the best ski pairs   in given conditions. The report describes cross-country skis and factors that   affect the performance of these skis. Moreover, this report presents the   testing procedure of the national team. The project provides a solution to   the problem in hand by developing a web service based on Django and Django   REST Framework and an iOS application to handle the user interaction. The app   was tested and approved by the waxers of the national team. To predict the   best performing skis in given conditions, the three Machine Learning   algorithms Support Vector Machine (SVM), Decision Tree, and Artificial Neural   Network (ANN) is implemented and evaluated. Experimental results indicate   that the ANN algorithm has better accuracy than the Decision Tree, and that   the SVM algorithms and that the SVM was performing slightly worse than the   other two, when applied on test data which is artificially generated based on   the experience of the national team. All three Machine Learning algorithms   perform better in terms of mean accuracy which is significantly higher   compared to the accuracy of a baseline algorithm. The report suggests that   the accuracy of the ANN algorithm is high enough to be useful for the   national team.