Classification of pitch types outside the laboratory or game environment provide benefits in designing and outlining training routines to reduce injury and improve performance. Current classification approaches are based on vision-based optical data, however these optical data is often not available in training sessions. The aim of the current study is to use machine learning algorithms to classify pitch types, i.e., fastball, change-up and curveball, based on pelvis and trunk IMU data in training sessions. A total of 406 successful pitches thrown by 19 pitchers were used to classify pitch types. Three conditions were tested and evaluated in a binary (fastball versus others) and multi-class (fastball versus change-up versus curveball) classification approach. Condition A included the direct output of the PitchPerfect software, whereas condition B included features normalized by fastball characteristics. Condition C is a combination of both condition A and B. The random forest algorithm demonstrated the best predictions in both the binary and multi-class classification approach based on the highest accuracy and F1 scores, i.e., the harmonic mean between sensitivity and precision. Therefore, the random forest algorithm is the best fit for classifying pitch types based on pelvis and trunk IMU data in training sessions. In addition, the performance of the classification algorithm improved when using a binary classification approach. There were no relevant improvements when using additional features. The random forest algorithm can directly be implemented in the PitchPerfect application. Pitchers can use the pitch type data to track and tune their performances, whereas coaches can use the data to design match and training routines. Future research should focus on larger datasets, i.e., more pitchers and pitches, to pre-classify pitchers with similar pitching characteristics to improve the classification algorithms.

Sports injury has long been causing concern to the athlete’s performance, financial aspect, and psychological impact. This is even more prominent in recent years as sports become more widely available o the mass population. Reducing the chances of athletes experiencing injuries not allows them to maintain optimal performance during training and competition but is also psychological benefits for the athletes. Amsterdam UMC has gathered weekly OSTRC data from 19 Waterpolo athletes over 109 weeks. By using this data, it is possible to build a model that can provide an indication of possible injury risks thus helping athletes in controlling possible injuries. Traditionally, statistical models are used for this kind of analysis. However, this often requires large amount of time and a-priori knowledge. Hence an injury risk classifier method based on longitudinal data (OSTRC) and machine learning algorithm was proposed in this study. The chosen ML algorithms for this study are ANN, LSTM, and Random Forrest. To investigate the importance of time dependency between data entries, sliding window, and forward chaining are used during data processing. All models trained with minimally processed data, sliding window data, and forward chaining data to investigate the impact of time dependency on model accuracy. The OSTRC data set is pre-processed, and SW and FC methods are applied. Data is then split into training and testing data set. Metric used to assess model performances are accuracy and confusion matrix. The results of this study show that both RF-SW and LSTM-SW produced an accuracy of 92.17% and 91.67% respectively. Accuracy of ANN on training data indicates adequate performance, but the confusion matrix indicates poor performance. Confusion of RF-SW and LSTM-SW show small variation due to differences in the test data set, which can be mitigated as more data are available. Inconclusion, the high level of accuracy from both RF-SW and LSTM-SW proves that it can be used to provide insight to athletes, helping them to reduce the chances of injury.