top of page

Machine Learning and how it helps make better investment decisions

Author: Richard Goh, CFO Smart Solutions

Machine Learning has come a long way since the 50s and with the increase in processing power with today's new systems, it's heavily being used to augment critical decisions being made in the FinTech industry. What it's able to do is provide insights without bias and do so in a relatively short time period where a slight difference in decision making can make a huge impact.

How do traders use to do it (before machine learning existed)

Before such tools existed in the finance industry, traders typically use their own intuition and past trading experiences in decision making. It has evolved from using basic excel sheets with cleverly constructed macros to help speed up some of the computations required. While trading can be fully automated and do exist, we're still at a stage where the final buy or sell decisions are still actions made by experienced individuals. The difference is the decision is now data driven and machine learning has helped allow the massive amount of data to be presented in a manner to allow efficient decision making. It's just like how commercial planes are all equipped with automatic pilot functions, yet the pilots are still required behind the flight yoke to make the critical and final decisions.

Is there such a thing as too much data?

There's a term called 'overfitting' where modelling of the data becomes a memorised and training approach rather than learning from the trends to generate useful results. While there are many ways to address overfitting - typically a combination of approaches are required. One preventive measure is to split the training and testing data and cross-validate the results to validate the results. Another approach is what's called 'ensembling', where multiple models are used for prediction - it's like having two different individuals to solve a fixed problem and validating if they come to the same conclusion and the approach taken.

What are the challenges Machine Learning for day trading

Machine Learning sometimes “look” at a longer term set of data to provide predictions but will always need to have the human element of experience to make a call. The intense day traders typically make profit from small changes in price, and it will make more sense for high-frequency trading with machines making the quick decisions to profit from the incremental changes.

Smart Solutions and our offering

RoboInvestor is built as a platform to aid investment professionals with a tool to crunch their data sets, simulate profit and loss with our backtesting tool and more importantly use only the data sets that impact investment decisions as technical and fundamental indicators in their model building. We embrace the domain knowledge of the person using our tool and allow for data and strategy unique to him/her to be catered for their use.

Are you a day trader ? RoboInvestor is capable of predicting daily signals with analysis computed on the granular dataset incorporated into the platform over a period as long as 20 years.



bottom of page