Veri madenciliği teknikleri ile girişim şirketlerinin başarısının değerlendirilmesi

Estimating the success of startups has attracted a lot of interest in the literature. In recent years, a lot of research has been carried out by researchers about the factors which influence startup businesses' successfulness and most corporations focused on developing various types of prediction models to accurately anticipate the fate of new businesses. In this work, we try to develop a reliable model for predicting startup businesses' success. Previous studies have concentrated on the accuracy of various algorithms without exploring the true influence of risk and success variables, or on understanding the effects of factors such as entrepreneur education, funding techniques, and timing on startup success or failure. In this thesis, we employ ensemble learning which is a method of machine learning to predict the success of startup businesses by using selected subsets of Crunchbase big data set. The goal is to use other variables from the dataset to distinguish between successes and failures. We include several factors that influence the startup success, and compare various prediction approaches, such as Extreme Gradient Boosting, Gradient Boosting, AdaBoost and Random Forest. The final result might assist Angel Investors and Venture Capitalists in developing more consistent and quantifiable startup portfolios. We randomly partition the data into two datasets throughout the estimate phase: training data and testing data. The objective is to use the benchmark approach (XGBoosting) to fit the model in the training data and then test the model's performance and the importance of the variables in the testing data. Keywords: Data mining; machine learning; ensemble learning; startup business, extreme gradient boosting