The successful application of machine learning algorithms in the financial sector has in more companies relying on stochastic models for the classification of financial distress. Researchers go to great lengths to make quantitative analysts get better accuracy than machine learning models. In this paper, we propose a technique for analyzing and classifying financial distress using the KNears Neigborn Classifier model, Random Forest, Naïve Bayes and the stacking ensemble. A collection of financial report data from the Indonesian Stock Exchange (IDX) containing financial reports is used as a basis for information as a feature of technical indicators. Selection of features based on their level of influence on the target output results with an accuracy rate of 98%. The stacking model of the neural network meta classifier ensemble is trained and tested on different multi-class ratio financial data sets combining technical indicator data of companies in Indonesia over time is novel. Our experiments show that in some scenarios, the ensemble stacking model performs better than the single model.
Ensemble, stacking, multi-class, classification