Motivation

Comparing the performance of many stocks can be time consuming, especially if you want to do it over and over again. With the help of R, Shiny and the availability of many open-source R packages, we developed a simple stock Analyzer- to do technical analysis and forecasting of a stock portfolio.

Approach and Design Framework

Our project will explore and develop an app for this using selected stocks from the US Stock Market.

Exploration: Exploratory Data Analysis explores the trends of stock prices and transaction volumes using TIME SERIES analysis. We select multiple stocks from different sectors and compare them within the same selected time period.

Technical Analysis: Many investors analyze stocks based on their fundamentals such as their revenue, valuation, or industry trends but fundamental factors aren’t always reflected in the market price. Technical analysis using charts help to identify trading signals and price patterns and provides as a window into market psychology to identify opportunities to profit. We built this with the core functions from TidyQuant[Ref.1] to support 1. Getting Financial Data from the web
2. Manipulating Financial Data
3. Performance Analysis and Portfolio analysis

Forecasting : As a final feature, we build a forecasting flow for a stock incorporating a dashboard to train several ML models( up to 8 ) and the associated visualisation and performance matrix of the trained models, as well as the forecasts comparison for the stock.

Key Output and Features

EDA For Single Stocks and Portfolio of Stocks

Exploratory Data Analysis implements a view of 4 equal sections, each showing a different types of chart.The charts can work with both single stocks, and portfolio of stocks.

Technical Analysis for a Single Stock:

1. Price Movement Candlesticks Chart
   
2. Stock Price and Trading Volume Line Charts/Bart Charts
   
3. Bollingor Charts + Moving averages (15-day, 50-day) Chart
4. Monthly and Yearly Returns

Technical Analysis for a Portfolio:

1. Closing prices
   
2. Moving Average
   
3. Bollinger Bands
4. Portfolio Monthly/Yearly Returns

Forecasting for A Single Stock

1. 8 ML (Machine Learning )models for forecasting
   
2. Allow user to tweak the ML model parameters training and forecast parameters
   
3. Visualisation of Forecast outcomes
   
4. Output in a table of the performance of ML training and Forecast outcomes

Learnings and Future Work

One of the most difficult challenge of building this application is designing the layout of all the components of a module. The application needs to find the right balance between showing too many visualizations, which can overwhelmed the users, and showing too little, which can make it harder to draw out meaningful insights.

To tackle this challenge, the use of more advanced Bootstrap features is needed, which allows the hiding and showing of different components, thus the users themselves can decide how many visualizations is suitable for them. The packages that enables these Bootstrap features are shinydashboard and shinydashboardplus.In implementing technical analysis, it was discovered that some tidy packages,lacked integration with plotly R package, which meant that interactivity was not possible with the chart. While convenient, there are some limitations to the timetk and modeltime packages, which can result in difficulties when trying to configure smaller details of the visualizations. In some cases, it might be better to modeltime purely for processing and training data purposes and plot the trained data from scratch using ggplot2 in order to get the desired visualization

Conclusion

In conclusion,leveraging the recenlty available R packages that have specialised/built in technical functions for Financial Analysis and Shiny, we can provide to the investor a easy interface to perform stock evaluation and analysis.

References

1.TidyQuant Project Website https://business-science.github.io/tidyquant/
2.Hyndman,R.J.,& Athanasopoulos,G(2021)Forecasting: principles and practice,3rd edition