Applying Supervised Machine Learning Models to Classify Day Traders Using the Traders’ Daily Trading Activity Data and the U.S. Stock Market Indices Data
Open AccessSome of the U.S. Stock Market trading firms (Firms) have Day Traders (Traders) that use several proprietary strategies that are based on passive - aggressive trading styles. The Traders are classified/grouped, based on their most recent trading profit results, to utilize the proprietary trading strategies that are based on their classifications. Based on their classification, the Profitable Traders are permitted to trade using the aggressive trading strategies and the not-Profitable Traders are permitted to trade using the conservative trading strategies. Reflected in the aggressive and conservative strategies are the amount of the Firms’ funds that the Traders can use to fund their trading activities, including the degrees of funds at risk. Some Firms utilize manual classifications model that is based on the most recent observed trading activities (Benchmark). To maximize the Traders’ profit and to minimize their loss by improving the traders’ classification, machine learning classification algorithm is used in this Praxis to classify the Traders as profitable or not profitable for the next trading day; the methodology uses a univariate “supervised” machine learning classification model, Random Forests (Model). Because the U.S. Stock Market (Market) is efficient, this Praxis does NOT attempt to predict the Market, instead, it learns the individual Traders’ attitude and aggressive interaction with the Market and attempts to predict/classify the Traders’ behavior for the next trading day. The Praxis used five months of a certain U.S. Stock Market Trading Firm (Firm) proprietary trading data (June through October 2017) to train and to test the Model. As a result, this applied machine learning classification Model, on average, outperformed the Benchmark by 16%, which could have saved the Firm $365k in opportunity loss during October 2017, the evaluated month. The Model also, on average, outperformed the enhanced Benchmark by 11%; the enhanced Benchmark is based on the 5 Day Moving Average of the most recent trader activities. Due to the sensitivity of the proprietary trading activity, only the Firm’s trading activity was available for this research.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.