Classification of Human Activity based on Sensor Accelerometer and Gyroscope Using Ensemble SVM method

Rapid technological development at this time is not only recognized by humans, now sensors embedded in smartphones can also recognize human activity using an accelerometer sensor and gyroscope sensor that has been embedded in it by producing hundreds or even thousands of records. accelerometer sensor and gyroscope sensor is one feature that serves to read the rate of change of acceleration from a smartphone but has a different function and requires data mining methods to group based on that output. Data mining methods that have better performance than other methods are Support Vector Machine (SVM) but are sensitive to parameter settings and sample training that cause undefined performance to overcome the shortcomings of the Support Vector Machine method by performing SVM ensembles, which are ensemble used is bagging. This research proposes the application of svm ensemble technique to perform human activity classification based on accelerometer sensor and gyroscope sensor. The results show that the best performance of SVM ensemble technique when comparing datasets with 70% training data and 30% test data with 99.1% accuracy, sensitivity 99.6% and specificity 98.7%.