This article presents a study that explores forecasting methods for multivariate time series data, which was collected from sensors monitoring CO2, temperature, and humidity. The article covers the preprocessing stages, such as dealing with missing values, data normalization, and organizing the time-series data into a suitable format for the model. This study aimed to evaluate Long Short-Term Memory (LSTM) networks, Convolutional Neural Networks (CNNs), Vector Autoregressive (VAR) models, Artificial Neural Networks (ANNs), and Random Forest performance in terms of forecasting different environmental dataset parameters. After implementing and testing fifteen different sensor forecast model combinations, it was concluded that the Long Short-Term Memory and Vector Autoregression models produced the most accurate results. The highest accuracy for all models was achieved when forecasting temperature data with CO2 and humidity as inputs. The least accurate models forecasted CO2 levels based on temperature and humidity.