A Contrastive Study of Machine Learning on Funding Evaluation Prediction

Chuqing Zhang,Han Zhang,Xiaoting Hu

doi:10.1109/access.2019.2927517

Chuqing Zhang, Han Zhang + Show 1 more

Open Access

https://doi.org/10.1109/access.2019.2927517

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2019
Citations: 5	License type: CC BY 4.0

Affiliation: Tsinghua University

Abstract

Winning a granting is critical in helping young and innovative firms to reduce financial burden, yet successfully get funding is not easy. Granting applicants are eager to find out the funding evaluator's decision pattern and fully prepare for the fund application. In such condition, supervised machine learning models seem to be a suitable tool. Based on nearly 5000 Beijing Innofund applicants, we find that supervised machine learning models, like support vector machines (SVM), K-nearest neighbors (KNN), decision tree, logistic regression, and Artificial Neutral Network (ANN) can produce both accurate and reasonably understandable funding prediction results with their average accuracy rate over 80%. Yet, the comparison results also reveal that the SVM model produces the most accurate forecasts in terms of average accuracy rate (86%) and F-score (82%). The findings indicate that SVM is an effective and reliable classification algorithm that can perform tasks well with small datasize. Based on the selected attributes and their weights, the funding applicants can get ready for the grants, by making up for the disadvantages and enhancing the advantages.

Highlights

Financial funding is a serial of special funds allocated by government
Based on nearly 5000 applicants of Beijing Innofund, we find that support vector machines (SVM) model produces the most accurate forecasts in terms of average accuracy rate (86%) and F-score (82%)
Logistic regression, decision tree, K-nearest neighbors (KNN) models are sensitive to unbalanced data, so they can hardly provide reliable results if the number of positive case and negative case are of big difference

Summary

Introduction

Financial funding is a serial of special funds allocated by government It plays an important role in reducing young and innovative firms or individual’s financial burden and supporting their innovative activities [29]. Supervised machine learning methods, like KNN, decision tree, logistic regression, ANN and SVM can improve decision-making quality and prediction accuracy by figuring out unknown dependencies or structures [2]. [12] compares credit rating analysis with an SVM and traditional neural networks in Taiwan and the US According to their experiments, SVM performs better than neural networks and logistic regression models. The results of [20] shows that the SVM provides the best accuracy for bankruptcy prediction among back-propagation neural network, radial basis function neural network, linear discriminant analysis, and naive Bayes classifier

Methods

Results

Conclusion