| Title |
Įmonės mokumo problemų identifikavimo sistemos tyrimas / |
| Translation of Title |
Research of company’s solvency problem identification system. |
| Authors |
Butkus, Mantas |
| Full Text |
|
| Pages |
48 |
| Keywords [eng] |
invoice payment ; machine learning ; classification |
| Abstract [eng] |
On time payment of invoices is an important factor that influences a company's solvency. Banks usually do not follow their customer payments until the customer pays his invoices. When the bank notices that his customer invoices are not paid usually it is too late and the company goes bankrupt. To avoid this it is important to identify companies that have solvency problems. In order to identify these companies company solvency prediction models are made. These systems alert the bank, that a company may have solvency problems in the near future. The aim of this work is to research a company's solvency problem identification system using not only regression but also classification models. Following tasks were created: 1. research company solvency detection systems that use only invoice information in other publications; 2. create a classification model for the current system; 3. create an adaptive company solvency problem detection system; 4. evaluate the quality of the system. Decision tree ensemble, support vector machine, Bayes and deep learning neural network models were used and analysed. It was determined that 5 to 6 features are sufficient for a successful performance of the system. The optimal threshold for the importance of the feature is 0.35. The best forecasting results were obtained using the decision tree ensemble and support vector machine methods, however the deep learning neural network model may also be used for invoice payment prediction. This project was also published at the E2TA conference in Kaunas University of Technology. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
Lithuanian |
| Publication date |
2018 |
