Abstract:
Prediction using time-series data is a vital part of machine learning because it keeps the temporal information of historical data for forecasting. Time series analysis is extensively used in all sectors wherever the data is populated and estimated based on timing such as seconds, minutes, hours, days, months, quarterly, half-yearly, and yearly. However, the model accuracy relies on the number of observations (data), consistency, and the consequence of data. The contribution of this paper is finding the trend of an educational institution enrollment in the upcoming year using the statistical machine learning mode1. Then, a detailed study has conducted to find the capability of the statistical model observed in various scenarios to handle time-series data. The study reveals the factors (Model fitness, Best forecasting duration, Impact of Train/Test ratio in precision) affecting the model accuracy of the statistical algorithms. This work also fulfills the research gap where less work has conducted in year-wise cyclic data without any trend. The methods used for this experiment are Auto-Regressive Integrated Moving Average (ARIMA) and Simple Exponential Smoothing (SES) technique. Finally, the two models are compared and the research objectives are discussed with the experimental result. Mean Absolute Percentage Error (MAPE) and Root Mean Squared Error (RMSE) metrics are used for assessing the model precision. The experimental result proves that the SES model provides better performance than ARIMA and both models are executed with their own merits and demerits.
Page(s):
5189-5200
DOI:
DOI not available
Published:
Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 14, Year: 2022
Keywords:
Simple exponential smoothing
,
Time Series
,
ARIMA
,
Student Enrollment Prediction
,
Factors Affecting the Model