The key to accuracy in supervised machine learning models is the creation of an adequate training dataset. Several datasets were tested and the optimal dataset chosen after benchmarking the accuracy of the model.
Stocksnips identifies key financial sentiment features based on n-grams and computes their weights based on an extremely fast machine learning (data mining) algorithm for solving multiclass classification problems from ultra large data sets.
Big Text Analytics and AI
Stocksnips reads millions of articles using Natural Language processing and extracts relevant financial news snippets. These are attributed to the right company and then scored by Machine Learning models that have been trained to deliver accuracy.
Automated (near) real-time news analysis further expands the universe of quantified data available to security analysts and enables a richer set of analysis for rapid and actionable changes to portfolios or trading strategies.
Stocksnips is a cloud based SAAS service with support for Android and iOS mobile devices. The back-end engine is deployed on Amazon Web Services ( AWS ) and uses a No SQL database. The architecture is scalable and can support millions of users.