Enhancing Snake Plant Disease Classification through CNN-Random Forest Integration
Deepak Banerjee, Vinay Kukreja, Samir Rana, Manika Manwal, Shanmugasundaram Hariharan
Abstract
This study proposes a robust classification strategy for accurately identifying several snake diseases of plant leaves. The model's performance in categorizing diverse types of diseases is thoroughly investigated using comprehensive evaluation criteria such as precision, recall, and F1-Score. The obtained results were impressive, with precision values that range from 69.77% for "Rust in Snake Plant" to 86.11% for "Root Rust and Snake Plants." The recall rates, which indicate the model's ability to detect instances properly, span from 73.60% of "Powdery Moisture with Snake Plant" to 80.40% of "Red Tree Spot and Snake Plant." Significantly, the F1-Score, which measures the balance of precision and recall, performs admirably, with values that vary between 73.50% for "Rust in Snake plant life" to 80.52% for "Root Rust and Snake Plants." The model's overall accuracy, as shown by support proportion, ranges from 0.88 to 0.93 throughout several illness categories. Furthermore, macroeconomic variables and weighted averages show a harmonious balance between accuracy and recollection, yielding an F1-Score of 76.61%. Surprisingly, the micro average replicates recall accuracy, and the F1-S at 76.51%. These data indicate the model's robustness, with substantial potential for early identification of diseases in snake plant leaves, leading to effective plant health control and supporting agricultural practices