Automatic Temperature Control Using Machine Intelligence

Abstract

: This project presents a machine learning solution for predicting indoor room temperature using the Random Forest Regress or algorithm. The model leverages a comprehensive dataset incorporating outdoor conditions (temperature, humidity, wind speed), building characteristics (room size, window count, insulation quality, sun exposure), and room usage patterns (occupancy, HVAC status, time of day). Categorical features, such as sun exposure and insulation quality, are preprocessed using Label Encoding and all numerical features are Standard Scaled to prepare the data for training. The trained model demonstrated excellent predictive performance, achieving an R² score of 0.93 and a Mean Absolute Error (MAE) of pm1.1 circa text{C} on the test set. Outdoor temperature and the status of the heating/AC systems were identified as the most influential factors. The resulting model and preprocessing artifacts are saved for deployment in a Stream-lit web application. This application provides an interactive tool for homeowners or facility managers to predict room climate, analyze thermal insights, and optimize HVAC settings for a comfortable range of 20 – 24 circa text{C} . The solution offers a high-accuracy, physics-informed approach to smart climate control.