Your typical investment research products would require users to download data via excel or an API followed by rigorous data cleaning with custom code in order to apply the latest machine learning models. This whole process would take numerous hours, packages and intricate steps only known by practitioners in the field.
OpenBB is bridging the gap for financial industry professionals by providing them the tools to be able to work with state-of-the-art machine learning with little to no coding experience. Industry professionals and technical practitioners are able to spend more time running experiments instead of troubleshooting implementations or reimplementing state-of-the-art Machine Learning architectures.
The Artificial Intelligence & Machine Learning (AI/ML) Toolkit provides practitioners with high-level components that can quickly provide state-of-the-art results, be it with with classical or deep learning models, while also providing researchers with low-level components that can be mixed, matched and fine tuned to build new approaches and custom tuned models.
Bring in multiple datasets, explore data with rich numerical metrics, create and add unique feature engineering and train machine learning models with unlimited external factors to see how underlying data may change future forecasting predictions and accuracy. Due to our open source nature, anything is possible from an end user perspective.
Furthermore, practitioners of all levels can experiment with applying data analysis techniques, cleaning, combinations and feature engineering for any sort of dataset found both within the terminal and external.
In order to do this, we selected to start integrating the Darts library, which is a leading open source time series machine learning library. Darts, whose development is led by Swiss Data & AI company Unit8, offers a large variety of forecasting models, from classical (such as ETS and Linear Regression) to state-of-the-art machine learning and deep learning (such as Neural Bayesian Estimation and Temporal Transformers). Many of the models can be trained on large datasets containing multiple series, allow the integration of external data, and can produce rich probabilistic forecasts. The underlying implementations are usually among the most performant around. The rich abstractions offered by Darts are reflected concisely and clearly in the AI/ML Toolkit, by leveraging the dynamism and the flexibility of the OpenBB terminal. The OpenBB development team have integrated the following models from Darts:
- Probabilistic Exponential Smoothing
- Theta Method
- Probabilistic Linear Regression (feat. Explainability)
- Regression (Feat. Explainability)
- Recurrent Neural Network (RNN, LSTM, GRU)
- Block Recurrent Neural Network (RNN, LSTM, GRU)
- Neural Bayesian Estimation
- Neural Hierarchical Interpolation
- Temporal Convolutional Neural Network
- Transformer Network
- Temporal Fusion Transformer Network
The AI/ML Toolkit also includes models from Nixtla’s StatsForecast library, which makes it very easy and convenient for analysts to have a broader access to fast and accurate econometric models.
- Autoselect (AutoARIMA, AutoETS, AutoCES, MSTL, and more)
- Auto Arima
- Auto Complex Exponential smoothing
- Auto Error, Trend, Seasonality
- Multiple Seasonalities and Trend using Loess (MSTL)
- Random Walk with Drift
- Seasonal Naive
All of the auto time series features fit different models and automatically select the best one (according to a certain metric based on historical backtesting), making predictive analytics an out-of-the-box feature.
Future iterations will provide more expandability into new domains of forecasting that will allow users greater ability to introduce new exotic types of data that might provide an edge regardless of time patterns or frequency.
With the addition of the AI/ML Toolkit within the OpenBB Terminal, the open source community finally has everything in one single place; financial data, extensive commands, and state-of-the-art machine learning models that just work right out of the box.
In a future post, we will dive more depth about the partnership between OpenBB and Nixtla, how they began contributing, and how two open-source companies are collaborating to raise the bar of quantitative finance. Don’t forget to subscribe to our newsletter to keep an eye out for this here.
Excited to get hands-on experience with our forecasting capabilities? Download the OpenBB Terminal by clicking here.