Multi-factor Stock Selection Engine using Machine Learning

Problem

A long-only strategy, with a mandate to outperform the S&P 1500, was using the traditional human-driven stock selection method to construct its portfolio. However, the human-driven process was riddled with biases and notions of the Portfolio Managers and analysts. In addition, the costs and time taken to evaluate 1500 stocks was a big drag on fund profitability.


Our Solution

We built an extensive bottom-up security selection engine, that relied on a scientific factor creation pipeline that was used to build 100s of factors (based on fundamentals, technicals, governance, estimates, etc.), along with a machine learning model to blend the various factors thus created into a composite score for each stock in the universe.


Impact

Our engine allowed the strategy to evaluate every single security in the benchmark, on an extensive set of factors, and with a consistency that had never been seen before. Moreover, the factor selection pipeline that we built has enabled the continuous improvement of the model through additional factors. The engine continues to generate market-beating model portfolios.


To know more about this case study and other examples in more detail, just send us a   Message