[Paper Review] A Data-Driven Framework for Identifying Investment Opportunities in Private Equity

Samantha Petersone, Alwin Tan+3 more|arXiv (Cornell University)|Apr 4, 2022

Private Equity and Venture Capital4 citations

TL;DR

This paper proposes a data-driven, explainable AI (XAI)-enabled framework for automating private equity (PE) investment opportunity screening by integrating financial, director, and deal data from UK-registered firms. Using multiple ML models and robust handling of data imbalance (via under-sampling and SMOTE), the framework achieves high predictive accuracy (F1 up to 0.99) and provides interpretable recommendations, significantly improving decision efficiency over manual methods.

ABSTRACT

The core activity of a Private Equity (PE) firm is to invest into companies in order to provide the investors with profit, usually within 4-7 years. To invest into a company or not is typically done manually by looking at various performance indicators of the company and then making a decision often based on instinct. This process is rather unmanageable given the large number of companies to potentially invest. Moreover, as more data about company performance indicators becomes available and the number of different indicators one may want to consider increases, manual crawling and assessment of investment opportunities becomes inefficient and ultimately impossible. To address these issues, this paper proposes a framework for automated data-driven screening of investment opportunities and thus the recommendation of businesses to invest in. The framework draws on data from several sources to assess the financial and managerial position of a company, and then uses an explainable artificial intelligence (XAI) engine to suggest investment recommendations. The robustness of the model is validated using different AI algorithms, class imbalance-handling methods, and features extracted from the available data sources.

Motivation & Objective

• To address the inefficiency and subjectivity of manual PE investment screening in the face of growing data volume and complexity.
• To develop a holistic, data-driven framework that integrates financial, director, and deal-level data for comprehensive company assessment.
• To validate the framework’s performance across diverse ML algorithms, class imbalance techniques, and feature sets to ensure robustness and reliability.
• To provide explainable AI outputs that support transparent, defensible investment decisions in high-risk PE contexts.
• To overcome data scarcity and imbalance issues common in PE by leveraging under-sampling and SMOTE techniques for improved model generalization.

Proposed method

• The framework integrates data from three sources: Unquote (private equity deal data), Fame (financial statements), and Companies House (director and company registration data).
• Exploratory data analysis (EDA) is conducted to understand data distribution, missingness, and feature relevance across the UK-registered firm dataset.
• Multiple machine learning models—Logistic Regression, Random Forest, XGBoost, SVM, KNN, and Decision Trees—are trained and compared using standard evaluation metrics.
• Class imbalance is addressed using under-sampling and SMOTE techniques to prevent model bias toward non-investment outcomes.
• Feature engineering extracts financial ratios, director experience metrics, and deal-specific indicators to enrich model inputs.
• Explainable AI (XAI) techniques are applied to interpret model predictions, ensuring transparency and trustworthiness in investment recommendations.

Experimental results

Research questions

• RQ1 Can a data-driven framework effectively automate the identification of promising private equity investment opportunities using multi-source data?
• RQ2 How do different machine learning models perform in predicting PE investment outcomes when trained on sparse, imbalanced UK firm data?
• RQ3 To what extent do class imbalance handling techniques (under-sampling and SMOTE) improve model performance and fairness in PE investment prediction?
• RQ4 Which combination of financial, director, and deal-level features yields the most accurate and robust investment recommendations?
• RQ5 How effective are explainable AI methods in providing interpretable and actionable insights for PE decision-makers?

Key findings

• The Random Forest and XGBoost models achieved the highest F1 scores of 0.99 on the hold-out set when using all features and SMOTE sampling, indicating strong predictive performance.
• The framework achieved an F1 score of 0.99 with the Random Forest model on the SMOTE-balanced dataset using all features, demonstrating high precision and recall in investment prediction.
• Under-sampling and SMOTE both improved model performance, with SMOTE yielding slightly better results on the test and hold-out sets, especially in recall and F1 metrics.
• The inclusion of director-level features (e.g., experience, tenure) significantly improved model performance, highlighting the importance of managerial quality in investment success.
• The XAI component successfully generated interpretable feature importance rankings, enabling PE firms to understand and validate model-driven recommendations.
• The model maintained high accuracy (up to 0.94) and AUC (0.94) on the hold-out set using all features and SMOTE, confirming robustness across different data splits.