Mining unstructured data for investment insights using Google Cloud Platform and machine learning
Pluribus Labs collects unstructured content from disparate sources including social media, corporate earnings calls and filings and mobile footfall data, and uses advanced quantitative research and machine learning to find insights and deliver investment advice to financial professionals around the world.
Pluribus Labs previously hosted its data on an alternative cloud provider, but found that it required too much infrastructure management and lacked sophisticated tools for mining information from such a broad range of unstructured data types. Pluribus Labs moved to Google Cloud Platform (GCP) because its suite of tools, including Google Cloud Pub/Sub, Google Cloud Bigtable and Google Cloud Dataflow, work together to offer powerful analytics capabilities. As part of its roadmap, the organization is particularly interested in using Google Cloud Machine Learning for advanced analytics challenges, such as gaining useful financial insights from Twitter.
“We chose Google Cloud Platform because it was the right solution for our business from a security, scalability and reliability perspective.” — Sebastien Astie, Chief Technology Officer, Pluribus Labs
Extracting useful information from social networks with Google Cloud Machine Learning
Pluribus Labs mines a wide variety of unstructured data for financial information. It searches Twitter and StockTwits, a social investing platform geared towards financial professionals, to derive investor sentiment on U.S. markets, industries and companies. It pores over the full transcripts of company earnings calls to provide an enriched understanding of companies’ future outcomes that financial statements alone can’t provide.
Using its own tools, it extracts information from unstructured data and puts it all into its data format. It uses Google Cloud Pub/Sub as the central means of sending the data to GCP tools, including Google Cloud Dataflow and Google Cloud Bigtable, which extract usable information from the raw data.
Moving ahead, Google Cloud Machine Learning can help perform sophisticated analytics tasks, such as separating spam from useful information in Twitter and StockTwits, and building pictures of investor sentiment about specific companies and broader economic sectors, such as consumer durables. This information and other insights generated by GCP are made available to investors via an API so the data can be applied to their own investment models. “Google Cloud Machine Learning is an extremely powerful tool for extracting useful information from unstructured data. We can provide investment managers with actionable insights quickly, at less cost and with more accuracy than if we used an alternative,” says Astie.
Reducing costs and speeding time to insight
With Google Cloud Platform, Pluribus Labs has reduced costs, provided insights to its customers more quickly and freed its engineers from managing infrastructure so they can focus on delivering better data to customers. When compared to its previous cloud provider, according to the customer, Google Cloud Platform is approximately 50 percent less expensive to operate. “We provide our customers with actionable insights based on unstructured, unconventional sources of financial information. We’re looking forward to offer even more sophisticated analytics based on what’s possible with Google Cloud Machine Learning and other tools such as Google Cloud Speech API and Google Cloud Vision API,” Astie says.