Machine learning for Stock/Crypto Trading

Harsh Ghodkar
4 min readMay 6, 2021


In this particular time period where software engineers are trying to Machine Learn everything, the majority of startups belong to the domain of ML. We, humans, are getting more dependant on machines as time flows by. But could Machines earn you money from stocks trading? Could they be trained to earn profits? Would you trust a machine to invest your money in a hope that you will get good returns? Could your trust issues handle that?

Well in this booming industry of Machine Learning, everything is possible and we are much nearer to the times where a bot could earn you money by investing and trading over it.

Looking at the current state of the stock markets, there are a ton of Mutual Funds, startups such as SmallCase who have hired experts to analyze the current market and invest the money of a customer at different places where maximum returns could be achieved. These experts before investing the money do check the plans of a company, look at the charts of the past years, consider hundreds of variables that could determine whether the company is the right place to invest and will it yield a good return. Thus we could conclude that these experts do a lot of homework over huge loads of data. What we are now considering is can this work be done by a machine?

Referring to one of the data analysis projects of my friend Vrajesh where he analyzed the data from various Social media platforms such as Twitter and Reddit and got to know about which Crypto/Stock is talked about in the market. Other than that project also includes analyzing the data of over 5000 cryptos and filtering the coins which have sufficient liquidity and have shown good returns in the past 24Hrs. As in crypto markets, anything can happen we could just rely on how popular the crypto is and how has its trading volume and liquidity grown over the past. The results of the analysis when viewed visually look like —

We have multiple APIs which allow us to gather enough data for our Machine to learn the scenario of Stock Markets. Analyzing that when crypto is falling and having a down trait but it has enough trading volume that one could invest in it. Learning the plans of a company from the scrapped data, analyzing that it might be a good option for a long-term investment plan.

What I am proposing is to build a model which will analyze data from all these APIs and will also place orders to buy/sell based on the analysis. Binance which is one of the top trading platforms has also an API that allows you to trade from your terminal.

Thus the model will place orders for you and trade when enough profit is made. Considering the current state, all the machine learning models which are trained, statistics say more than 60% of models fail when they are tested on Real World datasets. And as we are talking about making a model managing our money we require that the model should perform at its best accuracy.

The methods to train models are evolving at an exponential rate and they are getting much better with time. Thus in the future, we could see such models which would replace the current “Stocks Analyzer” and thus the mutual funds which are run by the experts currently would later be run by machines. Now consider that machines are trading against each other, and all the machines would be trained to get the best profit, some will make a profit by shorting the stocks while some will make the bull run. The bull and bears of the stock market would be converted to machines. The complete nature of the stock market might become binary and less volatile as we have seen that the sentiments and emotions of humans do play a big role in the rise and downfall of the market. Now we are talking about traders who don’t have an emotion, they will just be placing trades based on their training periods. It will be a fight of models vs models.

The future of the stock market is magical and we would have an awesome roller coaster ride observing it !!



Harsh Ghodkar

Coding Enthusiast | Machine Learning | Competitive Coder | Data Analyst | Full time Transponster