Optimizing Classification-Based Machine Learning Models for Bacterial Spot Disease Detection in Tomato Plants
Siya Sharma, Emily Broadhurst
Affiliation: Reservoir High School, Fulton, MD, USA
IJSCAR Vol. 3, Issue 4 (2026) · pp. 57–61
Abstract
On a worldwide scale, the farming industry loses $220 billion annually due to crop deaths. With approximately 160 tons of tomatoes consumed annually, about 50% of those crops are lost to diseases and pests. Our research aims to reduce the number of crops lost using early detection and machine learning. Modeling our data as a supervised learning problem, we can use images of tomato plant leaves to accurately detect bacterial spot disease, focusing on the impact that machine learning has on the reduction of food waste using accurate disease predictions. Bacterial spot disease was caused by the bacteria Xanthomonas perforans, most active humid, as tomato losses due to bacterial spot cost almost $900,000 in Florida were reported during the 2007-2008 growing season. (6) This paper explores how machine learning was used to detect bacterial spot disease, specifically in tomato plants. We hypothesized that tomato leaves infected with bacterial spot disease will show measurable differences in color and texture compared to healthy leaves, and these differences will become more pronounced as infection severity increases. We utilized and tested classifiers such as the Logistic Regression model, Ridge Classifier Model, Decision Tree Classifier alongside the Support Vector Classifier. Our Support Vector Classifier model had an accuracy score of 99.3%, along with a 99.4% recall score, which is more accurate than classifier models tested. This study indicates the usage of machine learning and early detection in tomato plants can strengthen sustainable agriculture, minimize food waste and improve food security.
Keywords: Plant disease detection, bacterial spot, tomato, image classification, machine learning, support vector classifier, convolutional neural network, PlantVillage, robustness to image changes