An Effective Motion Prediction Method for Chronic Diseases Based on PCA Data Fusion and MLP

Abstract

The statistical methods, such as paired t-test, odds ratio, and statistical values, have been widely used by medical analysis to measure the medical approach effectiveness. Those methods relied on massive hand-made data-collecting and hand-calculation, which put several limitations on the medical analysis: (1) small data volume, (2) unobtainable data feature, and (3) unpredictable data tendency. We present (1) using the three-axis acceleration sensor to collect data based on the big data-acquisition application platform. Also, (2) using the PCA method, which constructs a hyperplane, can adequately fuse the giant volume data feature. We are using MLP as the motion prediction model, which de-structures the data feature and then fits a parameter-based model to give the suitable motion value prediction scope. The proposed PCA-MLP method effectively moved the prediction close to the patient idea motion scope, which immensely decreases the short-handed problem on medical analysis and promotes the intellectualization development of the Big think on the medical analysis. We present a prediction framework from data evaluation to data prediction and trained the state-of-the-art model on our sensor dataset by resolving the sensor data.