Multisensor fusion sensor and improved Relief-F algorithms Based violence detection in schools
Narenthirakumar Appavu, C. Nelson Kennedy Babu · 2023
A widespread social problem is bullying in primary schools. Compared to school violence globally, primary school violence is seen as more harmful. Fusion and improved Relief-F algorithms are recommended as a multi-sensor-based method for early school violence detection. Data is collected by two motion sensors that play a major role in detecting violence in elementary school and in our daily life activities. A total of 7 types of violence related activities are checked. Time and frequency province characteristics, features are recovered and filtered by using the enhanced relief-F method. The authors then create a classifier with two levels. Random Forest is first level classifier which splits the jump function from others. In the previous paper the author used decision tree classifier but in this proposed work random forest classifier is used and the following level identifies the remaining 8 categories of functions using a radial basis function neural network. The recognition outcomes of the two sensors are then combined using a decision layer fusion method. Our research has shown that, on average, primary school violence can be recognized with an accuracy of 84.4%, whereas daily life can be recognized with an accuracy of 97.3%.