Final Year Project

Methodology

In this project, one or two industries of stocks will be chosen first. Industries include finance, energy, information technology and telecommunication and etc. Then, two or three categories of chosen industries will be selected. For example, in finance industry, there are several categories: Banks, insurance, investments and Assets Management. Two or three categories will be selected for exploration. After that, about three companies in the selected categories will be picked for investigation. Taking Banks categories as illustration, there are many bank companies categorized as banks such as HSBC, HANG SENG BANK and BANK Of E Asian. Data of two of them will be used for training the machine. The data collected can be the stock prices of selected companies of 5 years and the trading volume from 1-1-2013 to 30-12-2018 of those companies. The stock price is end-of-day price because it is hard to gain those instant price. The source of data, end-of-day price is from www.yahoo.com.hk.finance. Hence, the data collected is confidential and reliable. The testing environment is in windows 7 or windows 10. The platform of deep learning is Tensorflow invented by Google. The flexible architecture is the most important reason this project choose Tensorflow rather than other deep learning software library. This architecture allows user to deploy computation to one or more CPUs or GPUs in a desktop and server. According to Tensorflow official website, “TensorFlow was originally developed by researchers and engineers working on the Google Brain Team within Google's Machine Intelligence research organization for the purposes of conducting machine learning and deep neural networks research, but the system is general enough to be applicable in a wide variety of other domains as well. Besides, the population of using Tensorflow is increasing and hence the support for Tensorflow is plentiful and sufficient. The graph below shows the population over time of different deep learning libraries. The blue line is Tensorflow , and other colors represent different deep learning platform.

With use of Tensorflow library, an artificial neuron network can be built. In general, there are three main deep learning approaches widely used in different areas: convolutional neutral networks, Multilayer Perceptron Neural Network, Recurrent Neural Network will be used as different neuron network for training because these three neuron networks are quite different, it can be used to test which one is more efficient for stock prediction.