Embedded systems and artificial intelligence are changing modern agriculture as they work together to meet the growing need for accuracy, sustainability, and data-driven decision-making in climate-controlled farming. In this work, a low-cost IoT-enabled framework is developed that unifies real-time sensor data, including temperature, humidity, and pressure, with external weather inputs and crop-specific ideal conditions to facilitate crop label and farming resource prediction. The framework facilitates data collection, connects servers for data storage, enables crop type prediction, and also explains these predictions. Sensor data is carefully preprocessed through invalid reading removal, pressure fluctuation derivation, timestamp normalization, and resampling using linear interpolation and forward fill. These enriched sensor readings are aligned with crop condition profiles using Euclidean distance to compute crop cultivating resource usage targets based on weighted deviations. The system uses classification to predict optimal crop types and regression to estimate farm resource availability, leveraging models such as XGBoost, LightGBM, TabNet, and Feedforward Neural Networks. Among the tested models, XGBoost outperformed all others by achieving a perfect balance in classification evaluations (0.996 each), with an outstanding determination coefficient of 0.988 in the regression task. Explainable AI techniques like SHAP reveal key features and model decisions to ensure model transparency and practical deployment. This framework presents a scalable precision agriculture model with clear crop labeling, resource insights, and interpretable IoT-ML integration for smart farming decisions.