Rumour Analytics (Stock Exchange)
7th largest stock exchange
Our client is one of the largest stock exchanges in the world by market capitalization, with more than 5,500 companies are publicly listed on it. It is a major provider of integrated claims management services and claims review services.
Automated social media monitoring
Datametica has formulated an extensive approach to develop an automated social media monitoring solution for sites like Twitter and Facebook, and also on various news websites. This solution runs at a specified frequency and has automated their manual process of keeping a watch on social media. The end-to-end solution was developed on a distributed system - Hadoop to support the economies of scale.
We first find out if a listed company's name appears in the news article. If it does, then the data goes to the next stage, in which we use the Python natural language toolkit and remove stop words from the articles. We look at historical data for training the machine-learning algorithm.
Support Vector Machines, Naive Bayes and Neural Networks are used for training the statistical matching algorithm using the generated data. Detailed feedback is taken on the updated model and incorporated in a few iterations.
The solution provides an accuracy of identifying articles at 85% and avoiding false positives at 92% accuracy
Machine learning algorithms
By using machine learning algorithms, we shared combined information like retweet count, favorite count and if the Tweet has a link along with the text in the machine learning algorithm
We implemented an Alert system by interfacing with their social media dashboard through a database connection
The client’s surveillance team monitors news media for possible rumors about listed companies. Driven by highly competitive market conditions and the need to conduct fraud analysis, they intended to establish an automated solution for Social Media Analytics.
The goal was to tag news and social media content which can influence stock prices. Articles, which have material information or contain possible rumors about listed companies have to be identified in a timely fashion.