Stock Price Direction ML Classifier

Overview

This is a machine learning classification project in which the direction of Apple Inc.'s (AAPL) next-day stock price movement is predicted using information such such as Wikipedia page traffic, Google News traffic, and other technical indicators. The ML model outputs a binary outcome:

• 1 (the stock price increased)
• 0 (the stock price decreased/didn't change)

This project is an attempt to replicate the findings (in Chapter 2) of a whitepaper written by Bin Weng. The paper has been included in this repository, and can be accessed here.

Authors

• Adeet Patel
• Andrew Holpe (Richmond Quantitative Advisors)

Dataset

The aapl.xlsx file contains data for Apple ranging from July 1, 2015 to August 31, 2018.

Columns

• Open: The open stock price for the given day.
• Close: The close stock price for the given day.
• High: The highest stock price reached during the given day.
• Low: The lowest stock price reached during the given day.
• Change in Close: The previous day's close price subtracted from that of the given day.
• Gain: Equal to the absolute value of Change in Close if that value is positive, otherwise blank.
• Loss: Equal to the absolute value of Change in Close if that value is zero or negative, otherwise blank.
• Average Gain: The trailing 14-day average of the price gain.
• Average Loss: The trailing 14-day average of the price loss.
• RS: A momentum indicator, specifically the average gain divided by the average loss.
• Wiki Traffic: The amount of Wikipedia page traffic, calculated as the total of the page traffic for the following keywords:
  • AAPL
  • Apple Inc.
  • iPhone
  • iPad
  • MacBook
  • MacOS
• Wiki Traffic- 1 Day Lag: Equal to Wiki Traffic, but offset by 1 day. This is due to the fact that the Wiki traffic for a given day is not known until the next day. This feature is used in the model as it is more representative of the available data at a given point in time.
• PE Ratio: This feature is ignored.
• Wiki 5day disparity: The disparity for Wikipedia page traffic over a 5-day time period.
• Wiki Move: Boolean value; equal to 1 if the given day's Wiki traffic is larger than that of the previous day, and 0 otherwise.
• Wiki MA3 Move: The trailing 3-day average of the Wiki traffic.
• Wiki MA5 Move: The trailing 5-day average of the Wiki traffic.
• Wiki EMA5 Move: The exponential moving average of the Wiki traffic, calculated from Wiki MA5 Move.
• Wiki 5day Disparity Move: Boolean value; equal to 1 if the given day's Wiki 5-day disparity is larger than that of the previous day, and 0 otherwise.
• Goog Total: The amount of Google News traffic, calculated as the total of the traffic for the following keywords:
  • iPhone
  • iPad
  • MacBook
  • Apple Inc
  • iPod
  • Apple
• Change in Goog: The previous day's Google News traffic subtracted from that of the given day.
• Goog Gain: Equal to the absolute value of Change in Goog if that value is positive, otherwise blank.
• Goog Loss: Equal to the absolute value of Change in Goog if that value is zero or negative, otherwise blank.
• Goog Avg. Gain: The trailing 14-day average of the Google News gain.
• Goog Avg. Loss: The trailing 14-day average of the Google News loss.
• Goog RS: A momentum indicator, specifically the average Google gain divided by the average Google loss.
• Goog ROC: The rate of change for Google traffic.
• Goog MA3: The trailing 3-day average of the Google traffic.
• Goog MA5: The trailing 5-day average of the Google traffic.
• Goog EMA5: The exponential moving average of the Google traffic, calculated from the Goog MA5.
• Goog EMA5 Move: Boolean value; equal to 1 if the given day's Goog EMA5 is larger than that of the previous day, and 0 otherwise.
• Goog 3day Disparity: The disparity for Google News traffic over a 3-day time period.
• Goog 3day Disparity Move: Boolean value; equal to 1 if the given day's Goog 3day Disparity is larger than that of the previous day, and 0 otherwise.
• Goog ROC Move: Boolean value; equal to 1 if the given day's Goog ROC is larger than that of the previous day, and 0 otherwise.
• Goog RSI (14 days): The relative strength index for Google traffic, measured across a 14-day time period.
• Goog RSI Move: Boolean value; equal to 1 if the given day's Goog RSI is larger than that of the previous day, and 0 otherwise.
• Wiki 3day Disparity: The disparity for Wikipedia page traffic over a 3-day time period.
• Stochastic Oscillator (14 days): The stochastic oscillator for the low stock price, measured across a 14-day time period.
• Price RSI (14 days): The relative strength index of the stock price, measured across a 14-day time period.
• Price RSI Move: Boolean value; equal to 1 if the given day's Price RSI is larger than that of the previous day, and 0 otherwise.
• Google MA6: The trailing 6-day average of the Google traffic.
• Google_Move: Boolean value; equal to 1 if the given day's total Google traffic is larger than that of the previous day, and 0 otherwise.
• Target: Boolean value; equal to 1 if the given day's close stock price is larger than that of the previous day, and 0 otherwise. This value is predicted by the model.