An Empirical Comparison of Machine Learning Techniques: Focused on Crunchbase

 Glynnis Tan and Yesica Espinoza-Xitumul/ Dr. Byeonghwa Park

Kean University

tangly@kean.edu

Abstract

    

    According to Abraham Lincoln, the best way to predict the future is to create it. This is what Crunchbase strives to do. Crunchbase is a prospecting platform for dealmakers who want to search less and close more in terms of small businesses. With powerful features including personalized account recommendations, lists, territory preferences, intent signals, news notifications, and advanced search, Crunchbase aims to make it easy to target the right opportunities at the right time for small businesses. Since the prediction of the success or failure of a business venture has been something that people have struggled with for centuries, companies like Crunchbase’s best-in-class private company data offers insight into target companies’ teams, funding status, growth trends, tech stacks, web traffic, investments, and more, to personalize the outreach and increase engagement.  In this study, we will provide a comparison of different machine learning techniques for classification in startup’s outcomes (i.e. exit via IPOs, acquisition) from Crunchbase data. More than thousands of startups' data sets, including startup funding and status, can be found on Crunchbase. By using machine learning techniques, specifically logistic regression using Crunchbase data, we hope to predict the key factors needed for the success rate in a startup business. The machine learning techniques will be compared to see which technique is best to use and which one shows better and more accurate results. 

Introduction

        Startups are high-risk companies within their first stages of operations (Bento, 2017). To classify a business as a startup, innovation, the ability to grow, age, and profitability are often considered (Luc, et al. 2020). Over the years, there are many startups who’ve succeeded and obtained a space in the market, while others struggle to have the same results. The issue becomes how can startups know whether their business succeeds or fails? Are there any important factors that determine the success of a startup that they should be aware of? The ability of a startup company can meet its goals or predefined expectations may determine whether the startup is successful (Luc, et al 2020). Logistic Regression is a Machine Learning technique and can be defined as a modeling technique where the dependent variable takes binary values such as “0” and “1” (Bento, 2017). In this case, “0” means unsuccessful, and “1” is successful. Logistic Regression is used to predict the probability of a categorical dependent variable. Based on previous research, consistency with innovation, funding, and financial support are major factors in the success of startups (Okrah, et al., 2018).

Research Question: What variables play a key role in the success or failure of a startup company using the Logistic Regression technique?

Steps

STEP 1: WHAT IS A STARTUP?

    Startups are essentially, a high-risk company within its first stages of operations and usually related to products and services (Bento, 2017). Certain features that categorize a business as a startup include innovation, ability to grow, age, and profitability (Luc, T., Thanh, L., and Phung, N. 2020). 

STEP 2: IDENTIFYING THE FACTORS

    A successful startup depends on whether the business goals or predefined expectations can and are met (Luc, T., Thanh, L., and Phung, N. 2020). Success can be defined as “the measure of performance that occurs when businesses create value for their customers in a sustainable and cost-effective manner” (Luc, T., Thanh, L., and Phung, N. 2020). Survival is taken into account in the criteria of business success and is measured through performance and depends on the capability of the company to continue to operate (Luc, T., Thanh, L. and Phung, N. 2020). 

STEP 3: MACHINE LEARNING

    Competitive advantage takes a part in the success of a startup as it determines and secures them a place in the industry (Skawińska & Zalewski, 2020). Advantages primarily come from intangible factors such as distinctive capabilities and advantages of competence (Skawińska & Zalewski, 2020). 

STEP 4: LOGISTIC REGRESSION

    Research done by Okrah and Nepp, suggests that financial support is relevant for a startup to be innovative which in cause will affect their success, and is the main reason for bankruptcy and the exit of the startup from the market (Okrah & Nepp, 2018).

Methods

1. Defined a startup company and identified the factors that affect startup companies and their success.
2. Used Machine Learning technique, specifically Logistic Regression to predict which factors aided in the success of a startup company.
3. Built a Logistic Regression Model to identify the factors that affected startup companies. 

Results

    By applying the Logistic Regression technique to the Crunchbase dataset, we were able to eliminate any outliers that served no purpose in determining the success of a startup. We were able to create a model that determined the success and failure of a startup. Our model was able to achieve a 79.5% accuracy in predicting the success and failure of a startup company. 

Conclusion

Our study investigates, by implementing the Logistic Regression technique on the Crunchbase dataset, the algorithm that was used to predict the probability of a categorical dependent variable still needs to be explored.

For future research, we would like to see whether the accuracy of our model will increase or decrease if we tune our model. In addition, we would like to apply different machine learning techniques to compare with the logistic regression technique (i.e. Random Forest).   

References

Luc, T. T., Thanh, L. K. H., & Phung, N. T. K. (2020). Studying the Successor Startup Enterprises—A Case Study of Quang Binh 

Province, Vietnam. Open Journal of Business and Management, Volume 8(Number 4), 1426–1438. https://doi.org/10.4236/ojbm.2020.84091

Okrah, J., & Nepp, A. (2018). Factors Affecting Startup Innovation and Growth. Journal of Advanced Management Science, Volume 6(I), 

34–38. https://doi.org/10.18178/joams.6.1.34-38

Okrah, J., Nepp, A., & Agbozo, E. (2018). Exploring the factors of startup success and growth. The Business and Management Review, 

Volume 9(Number 3), 229–237. https://cberuk.com/cdn/conference_proceedings/2019-07-14-09-58-17-AM.pdf

Skawińska, E., & Zalewski, R. (2020). Success Factors of Startups in the EU—A Comparative Study. Sustainability, 12, 1–28. 

https://doi.org/10.3390/su12198200

Tong, Y., & Saladrigues Solé, R. (2019). An introduction to the study on start-up success. Start Up Notes, 1, 51–66. 

https://doi.org/10.21001/SUN.2019.1.04