Danijel Grah (2012) A system for graphology analysis support. EngD thesis.
In the diploma thesis we focused on the development of the system that belongs to software solutions in the field of graphology. In cooperation with an expert in the field of graphology we created a user friendly system that enables managing and saving the measurements. The Machine learning methods were tested on the data that we had at disposal. We decided to include the kNN method. A simple interface for entering any number of decision rules was also implemented. With the help of the expert graphologist 88 decision rules, which combine general predictions of personal features or professional types in the final part of the decision rules, were defined and entered. We tested the accuracy of the system with 10 manuscripts and compared measurements made with the original data. Larger deviations were noticed. In our opinion the cause for these mistakes is lack of professional knowledge in making the measurements. We established that we cannot uniquely conclude on a professional type based on prediction, nor can we thoroughly determine personal features and we think that the cause for inaccuracy lays in too generally defined final part of the decision rules. We have determined that we succeeded to create a system that substitutes standard tools which are used in a graphologists work. We also think that the system is suitable for demonstration purposes in teaching Graphology.
Actions (login required)