An object-oriented programming solution to compute average error between actual and predicted stock prices.

Before deploying a machine learning model for predicting real-time stock market data, it's necessary to know how accurate our predictions are over time by comparing our predictions with newly arriving real-time stock prices. This program was designed on the purpose of calculating average error of stock price in data streams.

Description

The idea was to use unordered map (Dictionary in Python Implementation) to store actual and predicted stock prices.

data structure:

key: stock id
value: 
	double-ended queue[0] for storing [hour, price] pair in actual.txt
	double-ended queue[1] for storing [hour, price] pair in predicted.txt

control flow chart:

• read in one window size of data from actual.txt, keep track of the breakline information, save into gap_line
• read in one window size of data from predicted.txt, keep track of the breakline information, save into gap_line
• calculate all the possible output inside this current window.(edge case: there is large hour gap inside actual.txt)
• write the calculated average error into output file.
• update window start/end information.
• pop out all not-in-the-window data.
• append gap_line information last.

Complexity:

Given that n represents number of hours, m represents number of stocks, k represents the window size: time complexity is O(nm), space complexity is O(km).

Feature:

• actual hours can start from any integer hour(>0). For instance, actual hour can be in range(144, 1440).
• predicted hours can be outside of actual hours. For instance, predicted hour can be in range(2, 6) while actual hour is ranging (2, 5).
• stock ids in each hour presented in actual.txt and predicted.txt can have different order.
• both actual and predicted data can have missing hours.

Dependencies

• import sys (standard Python module to obtain command line argument.)
• from collections import deque (for high performance Python data structure.)
• Python 2.7

Exectuion

To execute the code use the run.sh script.

./run.sh

Tests

6 Additional tests have been added in the insight_testsuite/tests folder. Those can be run by executing the run_tests.sh script.

cd ./insight_testsuite/
bash run_tests.sh

Or you can go to ./src/, run test_all.py

cd ./src/
python test_all.py