Time series based behavior pattern quantification analysis and prediction — A study on animal behavior

Abstract

The behavior pattern has regularity, reflecting the behavior feature and logic of the research object, and has a great influence on the prediction of the future state of the research object. However, the extant literature focuses on identification and classification of behavior pattern, lack of description and quantification research on behavior pattern. Behavior pattern quantified data can provide a good data foundation for behavior pattern prediction, further improving the accuracy of prediction. In this paper, we use the quantification algorithm based on Perceptually Important Point(PIP-QA) to analyze the time series, extract the hidden behavior pattern from the time series, and obtain the quantification description of the behavior pattern. A behavior pattern prediction model based on LSTM(BPPM) is also proposed to predict behavior pattern. Finally, the feeding behavior data of laying hen is used to carry out the experiment. The experimental results show the feasibility of the PIP-QA. And the BPPM model has good predictive ability and generalization ability.