Analyzing the Impact of the Threshold on Machine Learning Models for Credit Risk Prediction Using Business Intelligence

Aug 3, 2026

Upload by Aria

Making decisions from machine learning outputs requires users to understand how each prediction is produced. Predictive power alone is insufficient to build trust and acceptance of credit-risk models. This research proposes an approach to analyze how different probability thresholds affect binomial classification models in peer-to-peer lending. It defines measures for both global and local decision impact and demonstrates scenarios through a business intelligence application.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Explainability and interpretability of machine learning models

Explainable artificial intelligence (xAI)

Evaluating decision consequences globally and locally

Analyzing the Impact of the Threshold on Machine Learning Models for Credit Risk Prediction Using Business Intelligence Assoc. Prof. Dr. Yanka Aleksandrova University of Economics-Varna, Varna, Bulgaria [yalexandrova@ue-varna.bg](yalexandrova@ue-varna.bg) Chief Assist. Prof, Dr. Mariya Armyanova University of Economics-Varna, Varna, Bulgaria [armianova@ue-varna.bg](armianova@ue-varna.bg) Abstract Making decisions based on predictions generated from machine learning models requires users to have a clear understanding of the mechanisms and logic behind every prediction. From one side, business users must be convinced in the ability of the models to generate correct predictions. Predictive power, expressed by the different performance measures, is not sufficient for building trust and acceptance of machine learning models. Business users need additional techniques and tools for model interpretation and evaluation of the effects from decisions based on machine learning predictions. In this research paper we propose an approach for analyzing the impact of different thresholds for converting probabilities into predictions for binomial classification machine learning models applicable for credit risk prediction in Peer-to-Peer Lending platforms. We define a set of measures to explore global and local impact on decision-making process and present different scenarios in a Business Intelligence application built in an analytical and business intelligent platform. Based on the presented results we can draw conclusions that when choosing the best model and threshold users should consider a broad set of measures not only for model accuracy, but also should consider misclassification costs, financial results, asset portfolio structure, etc. Keywords: machine learning, business intelligence, credit risk prediction, explainable AI, artificial intelligence JEL Code: O33 DOI: 10.56065/IJUSV-ESS/2023.12.2.79 Introduction Machine learning models are used ubiquitously for decision making. One area of application is the risk assessment (Petrov et al., 2021) and particularly credit risk prediction, where a prediction must be made whether a loan applicant will repay the credit or not (Lohani et al., 2022),(Cetin et al., 2023),(Meshref, 2020) . Credit risk prediction is especially important in the field of Peer-to-Peer lending business models. Crowd lending market is one of the steady growing areas of alternative finance along with digital payments in recent year and especially after the COVID-19 pandemic (Cambrridge Center of Alternative Finance, 2023) . Unquestionably, Peer-to-Peer lending benefits investors as well as platforms, organizational, and individual borrowers. However, the unique characteristics of this industry and the ever-changing environment also present risks to this business model. An increase in the percentage of loans in default is one of the key dangers(KehayovaStoycheva et al., 2023) . Bad loans pose a major risk to borrowers, online P2P lending platforms, anda large number of new investors joining the sector. This establishes the critical significance of the procedure for evaluating borrowers and estimating the likelihood that the loan will be or not be repaid. When making decisions based on machine learning models, experts must have trust in models’accuracy and predictive power. The trustworthiness of a machine learning model depends heavily on the ability to interpret the model behavior, algorithm and discovered knowledge. The ability of a model to be explained and understandable to users is associated with its explainability and interpretability (Gall, 2018) (Miller, 2019) (Molnar, 2020) . The interpretability of the model determines how much a person can understand the reasons behind the generation of a specific prediction in a supervised learning (Miller, 2019) . The higher the degree of interpretation of a model, the easier it is to understand the reasons and mechanisms for making a decision or

FAQs

Why is predictive power alone not enough for trusting machine learning models in credit risk decisions?

Users need understanding of the logic behind predictions. Trust and acceptance require interpretation, evaluation, and decision-impact awareness beyond accuracy measures.

What does the proposed approach analyze in threshold-based credit risk prediction?

It examines how different thresholds convert predicted probabilities into binomial classification outcomes and how those choices affect decision-making.

Which kinds of measures should be considered when selecting the best model and threshold?

The results indicate considering multiple criteria, including misclassification costs, financial results, and asset portfolio structure, not only model accuracy.

Aug 3, 2026

Upload by Aria