User Manual
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 10
| Download | |
| Open PDF In Browser | View PDF |
CA400 User Manual Ben Kelly - 15337716 Supervisor - Martin Crane Abstract SmartPredict is a Web application trading platform which aims to help users make intelligent decisions about buying and selling of currencies (foreign exchange & cryptocurrency markets). This is achieved through machine learning and chart analysis. The application will provide the user with predictions for the market based on the knowledge it has acquired from past data. Recommending whether the user should buy, sell or hold their specific currency according to its current value. A user will be able to trade, view the current market sentiment and review the prediction of the machine learning algorithm all within the application. The following manual will be a guide to aid the user set-up, install and use the application on their own machine. Download and installation are not necessary to use the application, if the user does not wish to install the application locally it can be found online at http://kellyb45.pythonanywhere.com/smartpredict. Table of Contents 1. Installation………………………………………………………………………...2 1.1 Python & PIP.………………………………………………………….……………......2 1.2 Django & Library Prerequisites………………….……..…………………….……….2 2. General Use……………………………………...…………………...………….3 2.1 Registration……………………………………………………………………..…...… 3 2.2 Login……………………………………………………………………………….....… 3 2.3 About…………………………………………………………...…………….………… 4 2.4 Dashboard……………………………………………………...……………………… 4 2.5 Cryptocurrency Dashboard………………………………………………………...… 5 2.6 Foreign Exchange Dashboard……………………………………...……………….. 8 2.7 Logout………………………………………………………………...……………...… 9 Page 1 1 Installation 1.1 Python & PIP In order to install Smartpredict on the user’s local machine, some programmes must be installed. The following instructions will assume that the user is using MacOS or Ubuntu and has sudo privileges. To install Python, the user must go to https://www.python.org/downloads/ and download python 3.7. The user must then open the terminal and type the following command to install PIP. sudo easy_install pip Upon installation of PIP, the user will need to go to the following link https://gitlab.computing.dcu.ie/kellyb45/2019-ca400-kellyb45/tree/master/src in order to download the SmartPredict source code repository. 1.2 Django & Library Prerequisites The following libraries will need to be downloaded in order to run the app successfully: pip pip Pip Pip pip pip pip pip pip pip pip pip pip install install install install install install install install install install install install install django==2.1.5 matplotlib sklearn statsmodels keras numpy pandas pylab seaborn textblob tweepy oandapyv20 coinbase To run the server the user must be working in the directory in which the manage.py file is contained and run the following command. python3 manage.py runserver Page 2 2. General Use 2.1 Registration smartpredict/register If a user wishes to access SmartPredict they need only create their account. The user will not be able to access the app until they register. The user will be met with the following screen where they should enter a unique username and a password which meets the requirements specified and click ‘Sign Up’. 2.2 Login smartpredict/login Once registration is complete the user will be redirected to the login page where they should enter the details they just created and press ‘login’. Page 3 2.3 About smartpredict/about The about page is where the user can gain some more information about the app itself. Here the Functional Specification can be viewed as well as a link to the Gitlab repository. On the bottom of every page, a footer can be found which links to the creators LinkedIn profile. 2.4 Dashboard smartpredict/dashboard The Dashboard page is where the user can select which dashboard to view, The two markets which Smartpredict supports are Cryptocurrency and Foreign exchange. Page 4 2.5 Cryptocurrency Dashboard smartpredict/cryptodashboard The crypto dashboard is where the user can view data on two separate cryptocurrencies. The user is met with a table indicating the following: ● The current price. ● The expected closing price. ● The relative strength index. ● The sentiment analysis. This is where the user should focus his/her efforts as it contains the most amount of data pertaining to the specified currency. If the user is unfamiliar with RSI and Sentiment analysis the colour of the numbers should be noted. The number will glow: ● Green if the number is deemed to be positive/strong. ● Yellow if the number is deemed to be neutral. ● Red if the number is to be deemed negative not to be confused with a number with a value that is less than zero. These colours are simply to indicate performance, not the specific number shown. Graphs of each currency can also be seen, the data is graphed over the last 100 days of trading. The user should take note of the rise and dips of the graph this is where the currency has gained/lost value and may be a good time to buy/sell. The user can hover their mouse over each dot to get more data about each price point i.e. the price in euro and the date. Above the graph, we can see the output of the prediction algorithm as well as the RSI and the current price of the currency. The algorithm itself is Long Short Term Memory (LSTM) model trained on the past 2,000 daily price points (about five and a half years worth of data). If the user wishes to learn more about the machine learning aspect of SmartPredict they can consult the Technical Manual on the Gitlab repository. Page 5 For further analysis, a Moving Average Convergence/Divergence (MACD) graph is also displayed. This indicates trend momentum and can be useful when looking to invest. The graph is intended to act as a buy/sell trigger to possible investors. Users may buy the currency when the MACD (green line) crosses above the signal line (grey) and sell the currency when the MACD crosses below the signal line. Page 6 For some less technical analysis, the user can study the sentiment analysis doughnut chart displayed below. Accurately predicting currency prices solely on past data is a challenge, this is why sentiment analysis is utilised, it adds another dimension to the overall prediction. The system analyses Twitter to get a better understanding of what the overall opinion of the currency is. The latest 100 tweets are analysed to determine if they are positive, negative or neutral. The chart is then displayed based on these findings with an overall sentiment statement returned alongside it. Like all the graphs displayed on Smartpredict, it is interactive and a user needs only hover their mouse over the segments to analyse what percentage of tweets was deemed positive etc. Page 7 2.6 Foreign Exchange Dashboard smartpredict/forexdashboard The forex dashboard has a similar layout to the crypto dashboard. The user is again met with a table of currencies. The difference is that in foreign exchange terms we are now talking about currency pairs. The user is concurrently selling one currency while they buy the other. The user should be aware that the forex market is much more stable than the crypto market which is still extremely new and volatile. Just like the crypto dashboard the user should pay attention to the colour of some indicators, for example, the RSI in the right-most column below. In the forex dashboard, all the currencies are graphed together but more detail can be seen upon hovering the mouse over each price point. The users should pay attention to the right-hand side of the graph as these are the most recent price points and are usually more relevant when predicting future prices. Page 8 Sentiment Analysis is performed on Forex terms and is calculated and displayed just like the crypto dashboard. Again the user should make use of as much of this data as possible as it can give an insight into the current state of the market. 2.7 Logout To log out, the user need only click the ‘Logout’ option in the header of every page. Page 9
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.5 Linearized : Yes Producer : Skia/PDF m76 Page Count : 10EXIF Metadata provided by EXIF.tools