Virtually, everything including the output of prediction models is data. There are a lot of existing data vendors in the market specializing in a variety of events, and we have close relationship with them as a consumer of their data sets. We also build raw data sets from the scratch. We build higher level derived data sets using other data sets. These genuine new creation of data gives HAAFOR an edge compared to other places only relying on existing data vendors.
Even though our current focus is on stock market, prediction models are not just limited to stock prices. Our interest is open to every aspect of predicting the future. Quantitative analysis in the theory and data, diversified data sets, genuine ideas, developing power, and understanding of the real world, we need all of these to create good prediction models. From cleaning data values to pioneering new concepts in economy, each prediction models require a different set of skills and abilities from the researchers. Including but not limited to fundamental analysis, economic modeling, behavioral finance, human trader modeling, deep neural network, natural language processing, pattern recognition, data mining, and so on, prediction models are created by methods in so many different areas that we can not even guess what will be the next new idea that opens up a new area.
Pure prediction models are often not suitable for trading. So our researchers create the trading prediction models out of pure prediction models using deep knowledge of the prediction model space. Including advanced mathematical analysis, optimization to exhaustive search method, our trading prediction models are also made out of the collection of a huge knowledge base. We are generating trading signals with different time frame from mid-low frequency to high-frequency trading by utilising order book data. With those trading signals, we trade in global stock exchanges in US and Europe. We will soon add Asia-Pacific region.