This research will consider modern data analysis techniques to improve weather forecasting models that are specific for agricultural decision making. Bringing in a wide range of information, e.g., satellite imaging, soil moisture, historical crop yields, and weather stations' incoming data, helped us build and evaluate the predictive models, such as Random Forest, LSTM, SVM, and KNN, which state crucial agro-climatic variables. Observations showed that LSTM yielded the best results and eventually got the lowest MSE for temperature (0.021) and precipitation (0.010) but for soil moisture the LSTM yielded the lowest MSE (0.015). Random Forest got good performance in particular in temperature and soil moisture prediction. It's a region of MSE equals 0.025 and 0.018 correspondingly. Through the competition, SVM and KNN also managed to obtain fleshy accuracy, although their MSE values were slightly higher compared to those of LSTM and Random Forest. This result demonstrates that the deep learning and ensemble learning approaches are powerful enough to derive relationships within the agricultural datasets dataset and result in the improvement of forecast reliability that is essential for decision making by the agricultural sector. The study should be expanded on future research to include other datasets and conduct some verification studies to establish the kind of spreadability models across the variety of agricultural systems