Science fiction is now part of our everyday life as machine learning and artificial intelligence are being more and more embedded in our lives through functions such as visual and audio recognition performed by “digital intelligence” instead of human intelligence. I Know First applies machine learning in finance to predict the future state of the market, these predictions can be used for the development of deep learning trading strategies and result in investment portfolios with excellent returns and performance statistics as we will show below.

Deep Learning Systems

Explaining and predicting natural processes through mathematical notions derived by human analysis has of course been an invaluable tool in the development of technology and understanding of the world that surrounds us and natural processes such as seismic events, population growth, traffic jams, and weather are all examples of systems that can be predicted with reasonable accuracy with such models. Machine learning is a group of models for predicting these types of processes which sift through enormous quantities of historical data to learn the regularities within them without the need for humans to explicitly write the rules into the model and can be used to make successful predictions about the future state of these systems. Real world processes such as a person walking across a security camera screen may seem difficult for a machine to “understand”, but upon closer examination, such events present consistent patterns which a machine can quickly recognize given enough training without the need for the determination of explicit rules through human analysis.  Systems for determining and predicting patterns by letting the machine “write” a model for the observed data combine methods of computer science, statistics, and mathematics and are generally named “machine learning”. A very common approach to machine learning is deep learning which is composed of artificial neural networks: these networks like a brain (and in fact modeled after the brain) consist of layers of “neurons” or activation weights which process input data, effectively write the model by adjusting these weights to match past results, and generate predictions.

Machine Learning in Finance

It quickly becomes obvious that the uses of deep learning are many and very exciting of which one of the most interesting is the application of machine learning in finance. 40% of the world population is now online, and people use more than 2 billion smartphones every day. This is creating endless raw data for AI to process. This data encapsulates our thoughts, behaviors, interests, knowledge, etc. By analyzing this data systematic market patterns often emerge, which help us better understand the past and leverage this information to make realistic market forecasts and thus perform deep learning trading.

I Know First’s AI System for Stock Market Prediction

There are three different regimes the market can alternate from – positive feedback, negative feedback and randomness. Market analysis tries to recognize whether an investment is mean reverting or trending, and on what time scale. This can become very difficult, as these regimes may be present simultaneously on different time scales. Correctly analyzing this aspect is absolutely necessary for making accurate market predictions.

The I Know First market prediction system solves this difficulty by combining artificial intelligence, deep learning and genetic algorithms which provide distinctive insight into our comprehension of market dynamics and behavior. The algorithm has a built-in general mathematical framework that generates and verifies statistical hypotheses about stock price development. Machine learning tools such as artificial neural networks make this prediction system self-learning, and consistently determined to become more precise. New data is added daily to the 15-year database, where it runs a learning and prediction cycle that creates predictions for the short and long term.

I Know First’s Deep Learning Trading Strategies

Here we show the results of deep learning strategies that use the I Know First algorithm’s predictability filter to determine those assets that are predictable and have high expected returns and construct a portfolio of those assets. In these strategies, selection is focused on the S&P 500 stocks with the strongest three-month forecasts and high predictability levels. Initial portfolio size is 10 stocks for all strategies and positions are adjusted based on the daily updated forecasts, where shorter term signals are used as time elapses resulting in a maximum holding period of 63 trading days. As our algorithmic forecasts signal that price targets have been reached positions are closed and replaced by new ones based on the rankings on the respective trading date.

In the table below (click on the table to enlarge) performance numbers of two I Know First strategies (rows 1-2) versus the benchmark (SPY ETF, row 3) are presented for the period 08/2015-11/2017. As the constructed portfolios exhibit “path-dependence” and the performance numbers depend on the starting date average performance results for several starting dates are shown.

As can be sen in the table, I Know First’s systematically rebalanced deep learning trading strategies applying machine learning in finance register excellent performances: Total Returns of 53% and 60% versus the benchmark’s 33%, Beta of 0.13 and 0.16, an annualized Alpha ranging between 13% and 16%, Sharpe Ratios of 1.4 and 1.6, and Max Drawdowns below the benchmark’s. With an average holding period of around 18 trading days, this model delivers scalable strategies minimally affected by transaction and spread costs, that are suited for mutual funds and/or other financial products.

In the chart below the equity lines for the two strategies can be seen for the starting date 10/15/2015.

The charts give testament to the excellent statistics by showing stable and consistently growing returns over time.

Pilot with Institutional Investor

At I Know First we are currently testing the strategies shown above in collaboration with an institutional investor. This collaboration as of October 27, 2017 has resulted in the following equity lines.

The two selected strategies with Total Returns of 8.59% and 7.51% have significantly outperformed the benchmark (4.89%) in the analyzed time period, giving rise to annualized Alphas of around 30% and promising well for their future utilization in a fund which will allow retail clients to directly invest in an I Know First algorithm powered investment product. This fund will contribute to the growing applications of machine learning for trading and the permeation of successful artificial intelligence systems like I Know First’s algorithm for investment selection and monitoring.

Conclusions

Artificial intelligence is quickly expanding its influence in all types of daily tasks improving the efficiency of many processes and giving great promise for the future. Clearly, applications of machine learning in finance are also on the rise for investment selection and risk management tools. I Know First is on the forefront of this new field as the deep learning trading strategies based on its state of the art artificial intelligence system give rise to portfolios with excellent performance, vastly outperforming the benchmark and are well suited for the managment of mutual funds or other investment products.