Prediction of IPO performance from prospectus using multinomial logistic regression, a machine learning model

In this study, we investigated and determined the financial indicators that significantly affect the performance of initial public offerings (IPOs) by using the Multi-Logistic Regression (MLR) method based on several financial variables as independent variables extracted from prospectuses. The prospectus serves as an important source of information for potential investors as it significantly increases the likelihood of attracting their attention. A total of twelve characteristics in two segments, namely 'prospectus characteristics' and 'financial ratios', were used as the features to assess IPO performance in the Saudi stock market in three categories: BELOWAVERAGE, AVERAGE, and ABOVE AVERAGE performance. The accuracy, recall, precision, and F1 scores as well as the confusion matrix and AUC were used to evaluate our dataset. Python was used, as an Integrated Development Environment (IDE), to develop the model. Based on the results of the classification analysis, 71.4% of the predictions were accurate with an AUC of 0.71 for the ABOVE class. The most significant financial variable that affected IPO returns was the subscription quarter (SQ), followed by sector code (SC), and recent year net profit margin (NPM%). The MLR model had a higher level of accuracy when compared with other machine learning algorithms. By using the model developed here, investors can improve their ability to predict the direction of the return on their investment in an IPO, at least for the first month. A practical application of the MLR method is discussed in the paper, along with how it can be used to predict the probability of the performance of an IPO in the future.