Prediction of Temperature and Humidity at Telkom University Landmark Tower (TULT) using Regression Tree Ensembles and Internet of Things (IoT)

This paper addresses the significant impact of temperature and humidity changes on human comfort and health within indoor environments. The objective of this research is to develop a prediction model for air condition outside Telkom University Landmark Tower (TULT) using the Internet of Things (IoT) approach and Regression Tree Ensembles (RTE). In this study, the RTE technique is employed as the prediction method that combines multiple regression trees to enhance the accuracy of predictions. Historical data collected from the TULT building serves as the training dataset for the model. To assess the model’s performance, evaluation is conducted utilizing the R-squared ($R^{2}$) and Mean Absolute Error (MAE) methods to quantify the error. The evaluation on temperature measurements show that the model achieves a low MAE of 0.28608 in the test data. In the training data, the MAE is even lower, at 0.00397. For humidity measurements, the model exhibits a higher MAE of 7.0043 in the test data, indicating higher error rates in humidity predictions. Similarly, the MAE in the training data is 0.1086, showing that the model performs better on the data it was trained on but still experiences higher errors in the test data. In terms of $R^{2}$ values, the model achieves an $R^{2}$ of 0.3159 in the test data for temperature measurement, indicating that it can explain approximately 31.59% of the variance in temperature data. In the training data, the $R^{2}$ value is much higher at 0.9949. Interested readers can use the data from this website https://socj.telkomuniversity.ac.id/data/.