Financial Forecasting with Artificial Intelligence
Artificial Intelligence for Financial Forecasting markets are considered to be one of the leading indicators of the economy.
When markets begin to contract, the society braces for a slowdown in the economy or a possible recession; when markets expand, the economy builds a forward momentum.
There are many factors and forces that influence market movement. They are complex, high correlated, and sometimes difficult to measure. A challenging problem is how to make accurate financial predictions using this complex set of inputs under tight time constraints.
Matthew Butler, Master of E-Commerce, and Dr. Vlado Keselj, have developed Artificial Intelligence techniques that use machine learning to automatically learn patterns in the numerous financial inputs to predict performance of stock prices in a month to year term range. The experiments run on the stock market data have shown that their model outperforms the benchmark portfolio based on a stock market index, on investment return.
An interesting novelty is the use of text analytics methods to interpret non-numeric information provided in the annual reports submitted to the Securities and Exchange Commission (SEC) to the publicly-traded companies. Beside the numerical data in the reports, they exploited the narrative, textural components, which give insight into the opinions of the senior management team and provide direction of where they feel the company is going.
This information is not to be overlooked, but direct interpretation by analysts is very time-consuming, error-prone, and possibly subjective. Character n-gram based profiling, used by Dalhousie researchers, provided a way to capture this intangible information in an automatic and efficient way. The field expertise is provided by the external collaborator Dr. Vladimir Lucic from Barclays Capital, UK.
More information about this research and about other uses of Text Analytics in Authoship Attribution, Text Clustering, Spam Detection, Sentiment Analysis, Automatic Essay Grading, and Detection of Dementia of Alzheimer Type from Spontaneous Speech, can be found at the site of Dalhousie group for Natural Language Processing and Data Mining (DNLP).