G Graphology is the study of the relationship between handwriting characteristics and individual personality. conventional graphological analysis is subjective and highly dependent on expert interpretation, creating the need for a more objective and efficient automated system. this study develops a graphology-based personality inference system using the mobilenetv2 architecture with a transfer learning approach. a total of 100 handwriting images were processed using opencv to automatically generate labels for five graphological features: letter size, slant, spacing, baseline, and pen pressure. these features were then classified simultaneously using a multi-output mobilenetv2 model supported by data augmentation and class weights to address class imbalance. the classification results were mapped into personality profiles through a rule-based inference system consisting of 13 if–then rules derived from graphology literature. the experimental results achieved an overall average accuracy of 73%, representing an improvement of 15 percentage points compared with the previous study, which achieved an accuracy of 40%. the classification accuracies for each feature were 65% for letter size, 55% for slant, 90% for spacing, 75% for baseline, and 80% for pen pressure. the proposed system is capable of generating up to 108 personality profile combinations automatically through a transparent and scientifically traceable inference process.