TIME SERIES WITH LSTM IN MACHINE LEARNING

Abstract

Air pollution has numerous implications for agriculture, the economy, traffic accidents and health. In recent years, the fast development of business technology has been in the middle of severe environmental pollution. long faced with several environmental pollution issues, particulate (PM2.5) that is the subject of special attention & is wealthy in an exceedingly sizable amount of toxic and harmful substances. additionally, PM2.5 features a long continuance within the atmosphere and an extended transport distance, that the analysis of the distribution of PM2.5 is a vital drawback for predicting the standard of the air. Therefore, this project proposes a way supported a univariate remembering network (LSTM) to investigate the spatiotemporal characteristics of the distribution of PM2.5 so as to predict the air quality in many cities. within the prediction, information records of real words were collected and analyzed, and 3 measures of exactness (i.e. mean absolute error (MAE), mean root error (RMSE) and root error (MSE)) were wont to measure the performance of the tactic projected during this project. . For the analysis of the projected technique, the performance of the projected technique is compared to alternative machine learning strategies. The results of the sensible experiments show that the projected MAE, RMSE and MSE strategies square measure inferior to alternative machine learning strategies.