Soil-Specific Crop Recommendation with Explainable AI: The XGB-FuzzyIncNet Approach

ABSTRACT The economy of many world countries relied on agriculture. Its productivity is often affected by challenges like selecting suitable crops for a particular soil type. Farmers face difficulties in identifying the most profitable crops due to the diversity of soil properties globally. This issue is addressed in this work by developing an innovative crop recommendation system – Extreme Gradient Boosting Fuzzy Inception Network. It is formed by the combination of modules such as Extreme Gradient Boosting regression, improved inception v4 network, and fuzzy decision-making layer, which recommends the optimal crop to be grown in specific land by evaluating the various chemical and physical properties of the soil. In this design, the preprocessing phase cleans, reduces, and normalizes the raw data and prepares it for the classification process, in the classification phase the Extreme Gradient Boosting regression performs initial regression and finds significant features to produce a preliminary prediction, the improved inception v4 model captures hierarchical features from Extreme Gradient Boosting output to improve prediction and the final fuzzy layer fuzzifies the output of inception v4 into fuzzy linguistic variables to handle uncertainty and improve decision making. Additionally, explainable artificial intelligence is implemented to analyze the effect of every input parameter on the output. The results noted during the experimentations indicated that the proposed crop recommendation system exhibits its superiority over the existing techniques used for comparison by attaining superior values in all the evaluation measured used. The proposed model attained an accuracy of 98.78%, precision of 98.82%, recall of 98.76%, F1-score of 98.78%, specificity of 98.25%, MSE of 0.004, and RMSE of 0.02. The proposed crop recommendation system produced robust and accurate crop suggestions and enhanced the productivity of the particular crops on the appropriate soil types.