Rita Widyaningsih

CAPSTONE PROJECT Data Science for Accounting Portfolio Optimization Strategy using Phyton Regression Model 3 My Project Background (Business Problem) Problem How to Solve Invalid columns in datasets of the weight for stock picking concept Eliminate the data (not use for modeling) Assets diversification data is not ready Calculate Traynor and Sharpe Ratio and determine the portfolio can be well diversified or not. Small size and unclear source of datasets Unsolved. Try to eliminate the barrier by accurately preprocessing the data for a better model. 4 Method & Workflow Project Method used - Portfolio Optimization Strategy Theory - Data Science Steps using Python - Multivariate Linear Regression (based on theory) has been canceled and changed to Multiple Linear Regression (for better modelling) Workflow -. Import package Load Dataset and data pre-processing Define problem EDA and Feature Engineering Final Data Pre-Processing Modeling Evaluation Load and Predict New Data Conclusion and Recommendation 5 6 Result of Project Python offer a powerful tool for portfolio optimization and risk management. The regression model well surpasses in training and testing, although struggles to generalize same pattern to new data. Portfolio weight is vital for calculating excess return, invalid weight data source caused it should be 100% removed, indicating a reason of imperfect model Portfolio optimization relies on various indicators like macroeconomic fundamentals, financial statements, asset prices, and technical analysis, market value, diversification, dividend-paying stocks, invest in non-correlating assets (e.g., real estate, currency, bonds), put option, stop-loss order, and unforeseen risk. Invalid/insufficient above data makes bias and variance. 7 Recommendations: Python regression models provides objective and data-driven insights that guide investors towards optimal investment decisions. Variance between testing model and prediction model suggesting exploration of making other data models using new data of porthfolio • • • • Expand the Y-related dataset, including continuous and categorical data that influence Y(dependent variable). Ensure data validity, e.g., audited ROE data, listed stock prices. Increase the size of datasets and also diversity. Implement Polynomial Regression and/or Artificial Neural Networks due to complex datasets is significantly needed for portfolio optimization strategy. Use Ridge Regression and Lasso Regression to address inappropriate bias and variance in the modeling step.. 8 >>> Thank You!