Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Optimizing Private Equity Screening with Machine Learning: Applications in Nordic and UK SME Space
KTH, School of Electrical Engineering and Computer Science (EECS).
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
Optimering av Private Equity-screening med Maskininlärning : Tillämpningar för SMEs i Norden och Storbritannien (Swedish)
Abstract [en]

In this study, machine learning models are evaluated to determine their effectiveness in private equity (PE) screening for small and medium-sized enterprises (SMEs) in the Nordic and UK regions. The challenge with machine learning within screening for PE is the limited pool of companies that the PE firm is interested in having a dialog with, it is therefore important that the algorithm predict them as interesting, this is referred to as true positives. Minimizing false negatives is crucial, i.e. companies which the companies are interested in having a dialog with but the machine learning model predicts as not interesting, referred to as false negatives. A key consideration is ensuring that the proportion of true positives missed approaches zero as the confidence threshold decreases. Four machine learning models were evaluated: Histogram-based Gradient Boosting Classification Tree, Neural Network, Random Forest, and XGBoost. The Histogram-based Gradient Boosting Classification Tree model performed the best due to its high hit rate at a threshold where only 10% of true positives are missed. This model also has a favorable downward-sloping hit rate, which is important for real-world applications. The study concludes that machine learning can theoretically make the PE screening process more efficient for Nordic and UK SMEs. However, practical challenges remain in determining to what extent machine-learning models using only financial features can streamline PE screening. The framework used in this study provides a structured approach to evaluating similar problems in the future, with a focus on real-world characteristics at varying confidence levels.

Abstract [sv]

I denna studie utvärderas maskininlärningsmodeller för att bedöma deras effektivitet vid screening inom private equity (PE) för små och medelstora företag (SME) i Norden och Storbritannien. Utmaningen med maskininlärning inom PE-screening är den begränsade poolen av företag som PE-företaget är intresserat av att ha en dialog med, det är därav viktigt att algoritmen förutsäger de som intressanta, dessa kallas sanna positiva. Det är avgörande att minimera mängden falska negativa, det vill säga antalet företag som PE-bolaget är intresserade av att ha en dialog med men som maskininlärningsmodellen förutsäger som ointressanta, vilket kallas falska negativa. En nyckelfaktor är att säkerställa att andelen missade sanna positiva närmar sig noll när konfidensnivån minskar. Fyra maskininlärningsmodeller utvärderades: Histogram-basert Gradient Boosting Classification Tree, Neuralt nätverk, Random Forest och XGBoost. Histogram-basert Gradient Boosting Classification Tree-modellen presterade bäst på grund av dess höga träffsäkerhet vid en konfidensnivån där endast 10% av de sanna positiva missas. Denna modell har också en gynnsam nedåtsluttande träffsäkerhetskurva, vilket är viktigt för tillämpningar i verkligheten. Slutsatsen av studied är att maskininlärning teoretiskt sett kan göra PE-screeningsprocessen mer effektiv för nordiska och brittiska SMEs. Praktiska utmaningar kvarstår dock när det gäller att avgöra i vilken utsträckning maskininlärningsmodeller som endast använder finansiella egenskaper kan effektivisera PE-screening. Den ram som används i denna studie ger ett strukturerat angreppssätt för att utvärdera liknande problem i framtiden, med fokus på verkliga egenskaper vid varierande konfidensnivåer.

Place, publisher, year, edition, pages
2024. , p. 9
Series
TRITA-EECS-EX ; 2024:277
Keywords [en]
Extreme Gradient Boosting (XGBoost), Histogram-based Gradient Boosting, Imbalanced Dataset, Machine Learning, Neural Network, Nordic Region, Private Equity (PE), Random Forest, Resource Allocation, Screening, Small and Medium-sized Enterprises (SMEs), UK Market
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-350820OAI: oai:DiVA.org:kth-350820DiVA, id: diva2:1885071
Supervisors
Examiners
Available from: 2024-08-13 Created: 2024-07-21 Last updated: 2024-08-13Bibliographically approved

Open Access in DiVA

fulltext(415 kB)800 downloads
File information
File name FULLTEXT01.pdfFile size 415 kBChecksum SHA-512
b21a92d750387d0e41e30bee0a0486ffafb4e44e46be37dd32acc7adc4036417f5b63708c67a32b5e95dca15dbf63b51be0dafabdd709797131963e18376dabc
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 803 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 927 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf