For more information, please follow the Github Link:
ml-company-value
lamvytran • Updated Jan 9, 2024
Introduction
Astute investors need to know how to value companies accurately as the finance industry moves quickly. To determine if a company is overvalued or undervalued, this initiative makes use of big data sets including CRSP, Compustat, and some external market data sets. This will enable investors and decision-makers to gain practical insights into evaluating the position of a company in its industry.
Objective
By using other financial ratios and values as variables (long-term solvency ratios, asset management ratios, and profitability ratios), we hope to predict whether a company is overvalued or undervalued based on its P/E ratio. We achieve this by building models based on various machine-learning techniques.
The objective of this research is to determine which machine learning model produces the most significant forecast and to utilize that model to estimate our valuation as accurately as possible. Five distinct machine-learning algorithms were used in this research, and models based on each of them will be presented and discussed in the parts that follow.
Methodology
We will build a dumb model based on guesstimates and try different machine-learning techniques, such as:
- Random Forest
- Ridge and Lasso Regression
- XGBoost
- Classification Tree
- Neural Network
Then we compare the accuracy and evaluate some pros and cons of each model.
For more details, please have a look at this report: