Fuzzy Rule-based Classification Systems for the Gender Prediction from Handwriting

The handwriting is an object that can describe information about the author implicitly. For example, it is able to predict the gender. Recently, the gender prediction based on handwriting becomes an interesting research. Even in 2013, an competition for prediction gender from handwriting has been held by Kaggle. However, the accuracies of current approaches are relatively low. So, in this study, we attempt to implement Fuzzy Rule-Based Classification Systems (FRBCSs) for gender predictions from handwriting. Three stages are conducted to achieve the objective, as follows: defining some features based on Graphology Techniques (e.g., pressure, height, and margin on writing), collecting real datasets, processing on digital images (i.e., image segmentation, projection profiles, and margin calculation, etc.), and implementing FRBCSs. The implemented algorithm based on FRBCSs in this research is Chi’s Algorithm, which is a method based on Fuzzy Logic for classification tasks. Moreover, some experiments and analysis, involving 75 respondents consisting of 36 males and 39 females, have been done to validate the proposed model. From the simulations, the classification rate obtained is 76%. Besides improving the accuracy rate, the proposed model can provide an understandable model by utilizing fuzzy rule-based systems.