Assessing Children’s Fine Motor Skills With Sensor-Augmented Toys: Machine Learning Approach

Background Approximately 5%-10% of elementary school children show delayed development fine motor skills. To address these problems, detection is required. Current assessment tools are time-consuming, require a trained supervisor, and not motivating for children. Sensor-augmented toys machine learning have been presented as possible solutions to this problem. Objective This study examines whether sensor-augmented can be used assess children’s The objectives were (1) predict the outcome skill part Movement Assessment Battery Children Second Edition (fine MABC-2) (2) influence classification model, game, type data, level difficulty game on prediction. Methods in (n=95, age 7.8 [SD 0.7] years) performed MABC-2 played 2 games with toy called “Futuro Cube.” “roadrunner” focused speed while “maze” precision. Each had several levels difficulty. While playing, both sensor data collected. Four supervised classifiers outcome: k-nearest neighbor (KNN), logistic regression (LR), decision tree (DT), support vector (SVM). First, we compared performances classifiers. Subsequently, types classifier that best accuracy F1 score. For all statistical tests, ?=.05. Results highest achieved mean (0.76) was DT obtained from playing easiest hardest roadrunner game. Significant differences performance found scores between maze (DT, P=.03; KNN, P=.01; LR, P=.02; SVM, P=.04). No significant performing other 3 (DT vs P=.42; P=.35; P=.08) P=.15; P=.62; P=.26). only (combination level) significantly better than combination middle (P=.046). Conclusions results our efficiently school. Selecting (focusing or precision) (sensor data) more important determining selecting