The development of a prototype chatbot for automatic question answering and PM 2.5 prediction

Abstract

The objectives of this research are: 1) to study PM 2.5 dust data and its effects, 2) to compare the efficiency of PM2.5 dust forecasting using the forecasting error method, and 3) to develop the chatbots for answer questions about PM 2.5 dust, the effects of PM2.5 dust, and forecast values. In this research, three methods of forecasting error were used: global constant mean model, exponential weighted moving average, and adaptive-response-rate single exponential smoothing; by comparing PM 2.5 dust data and PM 2.5 dust forecast data. By finding the most efficient way is to give the lowest mean of the root mean square error (RMSE). From the experiment, it is found that the exponential weighted moving average method gives the best performance with an alpha of 0.1, a mean square error (MSE) of 5.45, an RMSE is 2.33, the mean absolute deviation (MAD) is 1.79, and the mean absolute percent error (MAPE) is 22.05%.