Development of Financial Distress Prediction Model for Non-Manufacturing Firms in Indonesia Using Support Vector Machine, K-Nearest Neighbour and Logistic Regression

Christian, Theodoros (2020) Development of Financial Distress Prediction Model for Non-Manufacturing Firms in Indonesia Using Support Vector Machine, K-Nearest Neighbour and Logistic Regression. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
02411640000072-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (1MB) | Request a copy

Abstract

Identifying firm’s financial health performance when in distress condition is important before the bankruptcy. One of the tools is to use financial distress prediction model to provide early warning of corporate failure. In Indonesia, the development of financial distress model mainly focuses on manufacturing firms. While there are rarely models developed specifically to non-manufacturing companies. This study applies three method, namely Support Vector Machine and K-Nearest Neighbour as the machine learning algorithms, and Logistic Regression as the statistical causal model to build financial prediction model Indonesian non-manufacturing firms. These methods are chosen since they less vulnerable to statistical assumptions and can construct FDP models for more complex data context. The data which used to construct the models are consist of 136 healthy firms and 42 distress firms. The combination of feature set from accounting and market perspective are used to build financial distress prediction model. The empirical result shows the best performance would be achieved by all of the algorithm when using the feature set of combination between market and accounting variable. Majority of the developed models would improve the performance from previous existing model. The prediction of Indonesian non-manufacturing firms in the period of 2019 shows the all of the model which developed in this research statistically significant to outperform Altman Z-Score and Support Vector Machine can reach the highest F1-Score. On the other hand, although the models and Distance to Default shows that there is not statistically difference for the overall performance, all of the developed model produced higher accuracy, precision, recall and F1-Score. Since there is no significant difference except for LR model 3, it is strongly recommended to use all of well perform algorithm if possible, to compare the result between one model to the other, and decide the firm’s financial condition based on the majority of the model’s prediction result.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Financial Distress Prediction, Bankruptcy, Support Vector Machine, K-Nearest Neighbour, Logistic Regression, Altman Z-Score, Distance to Default
Subjects: H Social Sciences > HB Economic Theory > Economic forecasting--Mathematical models.
H Social Sciences > HG Finance > HG4529 Investment analysis
Divisions: Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User: Theodoros Christian
Date Deposited: 21 Aug 2020 06:52
Last Modified: 21 Aug 2020 06:52
URI: https://repository.its.ac.id/id/eprint/79157

Actions (login required)

View Item View Item