See how AI analysts use artificial intelligence to ride the sea of ​​business

In the bustling world of financial services, the market continues to fluctuate and the profit figure is king. An investment company has targeted the "in the word line" of news reports, using artificial intelligence to help themselves, making full use of the vast amount of qualitative information outside the world, as if the current market halfway to kill a bit of gold.

Image from: Lorenzo Cafaro

Toronto-based Triumph Asset Management (reformly reorganized as Amadeus Investment Partners) is using deep learning to explore in this area. They analyze thousands of news articles on a daily basis to better predict market direction and make trading decisions.

Old school practice

Over the years, analysts have grouped relevant news articles, identified trends about specific company reports, communicated with traders, and responded to the market.

This process is quite time consuming, and many opportunities may be missed during the period, as the analysis of the article is only a small part of the overall, Triumph data scientist Andrew Tan said.

To accommodate the growing demand for data, the company turned to AI.

Tan said, “We believe that through deep learning, using its speed and accuracy, we can improve the overall analysis and overall workflow of the news. This in turn can lead to better results and overall performance.”

Using artificial intelligence as an analyst

Using GPU and CUDA Deep Neural Network (cuDNN) libraries, Triumph's data scientists provide news from proprietary databases to deep learning systems. The machine is trained to parse an article every three milliseconds so that hundreds of thousands of articles can be processed every day, something that was previously impossible.

The system recognizes hundreds of keywords in an article. An unsupervised learning algorithm called GloVe assigns a value to each keyword, and the rest of the model can interpret and use it.

The deep learning system will eventually lead to three outcomes: linking the article to the appropriate stock and company; identifying each article, giving a positive, neutral, and negative biased rating; and evaluating the news affecting the market possibility.

In the current period, all kinds of "fake news" have been able to penetrate the traditional news circle, and the company's data scientists use specific keywords and trusted news sources to improve the reliability of the system.

Although the system is still in its preliminary testing phase, Tan said that given that human analysts themselves would disagree on the suitability of propensity ratings, it is quite encouraging to initially achieve 76% accuracy.

He said: "This system is not perfect, but we will continue to build on this."

For more information on the application of artificial intelligence in the financial industry, please stay tuned to GTC China 2017 , the GPU Developer Festival, which will be held at the Beijing International Hotel Conference Center on September 25-27, 2017 . At that time, industry experts from Baidu, Ant Financial, Ping An Technology, Percentage, Shanghai Clearing House and Shenlan Technology will gather at AI and the financial sub-station, and take us to discuss in depth how artificial intelligence can help the innovation of the financial industry.