In ML Alpha, we aim to empower a thriving community of Data Scientists to revolutionize investing. To achieve this, we equip our community with our Machine Learning-ready Datasets, the Data Science Portfolios, and the Portfolio Marketplace. Learn how to make the most out of these features of our platform through this article.
Machine Learning Datasets
After registering on our website you can download our ML-ready datasets and start exploring Machine Learning algorithms for predicting rankings of future winning stocks (i.e. AI Scores) in various time horizons. The three defined time horizons for the targets are 2 Weeks, 1 Month and 1 Quarter.
We also provide three distinct datasets:
The “Training” Dataset, which contains historical market data alongside the ML targets for supervised learning.
The “Test” Dataset, similarly containing new (i.e. unseen) historical data but without the ML targets. This is designed for you to test your algorithm on unseen data. Of course you can already reserve some “unseen” data from your Training Dataset if you wish.
The “Predictions” Dataset, which contains current market data ready for new predictions without the ML targets (as they do not exist yet). This dataset is the one used to predict your “masked” AI Scores. You will submit this dataset to your ML Alpha Portfolio and the AI Scores provided by your algorithm(s) will determine the stocks that your portfolio will be made of.
You can download all these datasets after toggling to the Advanced Studio option of our Data Science Studio section (see below)
Please note that all our datasets contain “masked” historical and present features and targets. This means that the stock ticker as well as the timestamp of the features and targets are hidden from the user. Furthermore all the features are “normalized” and “anonymized” to avoid revealing which company is behind each datapoint. The idea is that Data Scientist make predictions purely based on the available data eliminating any possible opinion/bias about a given company or sector, etc.
Data Science Portfolios
The Data Science Portfolios enable you to transform your “masked” predictions into a selection of stocks that the algorithm predicts will perform best within the chosen time horizon. They also facilitate tracking the historical performance of investments based on your predictions. Additionally, the portfolio can be shared in the Portfolio Marketplace for further monetization.
To create a ‘Data Science Portfolio,’ follow the steps below:
1: Log in to ML Alpha
2: Navigate to “My Portfolios” and click on “Create New Portfolio”:
3: Give the new portfolio a name and click on “Data Science Portfolio” to create it.
4: You will then have to select the desired update frequency of your predictions (and hence your portfolio rebalancing frequency). We strongly recommend that the update frequency matches the time horizon of the targets you select during your algorithm training. For example, if you select target_1M (Monthly Predictions), you shall upload your predictions every month. This ensures that your portfolio composition closely follows the predictions of your algorithms.
5: After confirming you will be able to see the new portfolio alongside your existing portfolios.
6: Accessing your Data Science portfolio you can take note of the Portfolio ID, which you will need to used for submitting predictions
Congratulations on creating your first Data Science Portfolio!
Now you have learnt how to get our ML-ready Datasets and to create your Data Science Portfolio. But how can you submit your AI Scores/predictions? and how can you see which stocks are predicted to perform best over the chosen time horizon?
In order for the Data Scientist to craft their Portfolios purely based on their algorithm(s) predictions, the only way of updating a Data Science Portfolio is by submitting the predicted “masked” AI Scores to their portfolios via our APIs. For guidance on this go to the Advanced Studio option of our Data Science Studio section.
IMPORTANT: to submit your predictions you will need your API key. You can obtain it as follows:
1: Navigate to your profile by clicking on your user icon in the top-right corner:
2: Click on the ‘Rotate or Create’ button located beneath the API Key section:
Copy the API Key, and ensure not to share it with anyone or add it to Git or any other social platform. In case your key is lost or accidentally shared, you can still rotate it to invalidate the previous one.
We provide a Jupyter Notebook example on training and submitting predictions to ML Alpha; check it out here.
Once you have created your portfolio and started rocking it with your ai-based scores you can decide to take part in our Portfolio Marketplace.
Portfolio Marketplace
The Portfolio Marketplace offers Data Scientist (as well as other ML Alpha users) the possibility to monetize their hard research work.
Thanks to the Portfolio Marketplace other ML Alpha users will be able to see the performance of your portfolio without getting access to the actual portfolio composition. If they are interested in getting access to the composition and updates of your portfolio they can subscribe it for a monthly fee that will be set by you.
At the moment all portfolios are free to subscribe but soon you will be able to set a price for people to see your predictions and portfolio composition.
To bring your portfolio to the Marketplace you just need to toggle the option inside the portfolio.
We believe that with these three features of our platform we can empower a community of Data Scientist to revolutionize and democratize the access to AI-based investing!
For any questions, doubts, or requests, feel free to join our community on our Discord Server.