Advancing Dermatological Image Classification: GLCM-Based Machine Learning Insights

The prospects to improve skin illness via the utilization of artificial intelligence algorithms is what renders this study economically important. Machine learning may assist physicians detect people quicker and more accurately. The effective identification of skin disorders using machine learning could result in the development of large and readily available digital tests. A model was used in the present study to analyze the HAM 10000 data. Two hundred images in total were chosen at random; one hundred showed dermatofibroma diseases, whereas the other hundred displayed benign keratosis. Subsequently, these images were resized to prepare for additional examination. The statistical features of the gray level co-occurrence matrix were calculated from the image dataset by changing the distances 0, 5, 10, 15 and angles 0°, 45°, 90°, 135°. Five different machine learning models were subsequently trained and assessed based on these features. The study shows that the logistic regression model accurately detects and classifies various skin diseases. The logistic regression model showed exceptional performance, exceeding the expected results in terms of accuracy 91.50%, sensitivity 93.00%, and F1-score 91.36. The results of the study were most favorable when using an angle measurement of 135°.