Using Weka API for creating a custom classification application

Abstract

In real world applications, classification models can be built by repeatedly going through the steps of data preprocessing, building classification models with different dedicated algorithms, testing the resulted models, until building a good enough classification model is achieved. The model built in this way can be used in order to perform predictions on new data, with a degree of trust resulted by testing the model. All the operations described above can be done with Weka, which is a very powerful machine learning tool. The authors consider that after going through the previous steps, for many real world applications, it would be beneficial to go through a supplementary step, of developing a custom application for data classification, specific to the realized study. The proposed application will use the Weka API, which will allow it to rebuild the selected classification model, save and load it in/from a binary file and then make predictions with the help of a user interface perfectly adapted to the data set used. The proposed application makes it easier to make predictions for a final beneficiary. The paper presents how such a custom classification application can be developed, using the Weka API and Java programming language. In order to validate the proposed solution, the authors have built a custom application for the classification of data of patients who may have heart diseases, starting from the Cleveland heart disease data set, obtained from the UCI Machine Learning Repository. The proposed application allows saving and loading the classification model through serialization and deserialization, so that building it in order to make predictions is necessary only once.