Udacity-Capstone-Project

Predicting stock prices

To run the project:

First you will need to install the dependencies in case you don't have them. You can do this by running:

pip install -r requirements.txt

Start the server:

cd into the root directory and run:

python application.py

then you have two choices to use the application. By using the application routes, or by using the GUI on the web application.

Using the application routes:

Train a model on a stock:

You can train a model to make predictions for a stock by simply calling:

GET http://localhost:5001/train/<ticker>

If for example you wanted to train a model for the Google stock, you would call:

GET http://localhost:5001/train/GOOG

This will train a model on the Google stock, to predict for the next 7 days. However you can specify dates for training, or a number of days ahead to train.

parameters:

pfrom: date in the format yyyy-mm-dd. Must go along with the puntil parameter
puntil: date in the format yyyy-mm-dd. Must go along with the pfrom parameter
ndays: days ahead from now to predict for. Must go alone.

Example calls:

GET http://localhost:5001/train/GOOG?ndays=10

Will train a model to predict for the next ten days

GET http://localhost:5001/train/GOOG?pfrom=2012-02-02&puntil=2012-02-09

Will train a model to make predictions for the week from Feb 2, 2012 to Feb 9, 2012.

Response:

When you call any of these routes, you will get back a key in the format: <ticker>_<days_ahead>_<begin_date>_<end_date>

Examples:

GET http://localhost:5001/train/goog

Response: {"key": "goog_7_*_*"}

GET http://localhost:5001/train/goog?pfrom=2012-02-02&puntil=2012-02-09

Response: {"key": "goog_7_2012-02-02_2012-02-09"}

Getting your results:

To get the prediction results, you should call the route /predict with the key that came as a response to your previous requests. Like this:

GET http://localhost:5001/predict?key=goog_7_*_*

And you would then get back a result like this:

[
  930.4221, 
  930.4581000000002, 
  930.3882000000001, 
  925.3549000000003, 
  927.3287, 
  929.2324, 
  927.2102000000002
]

Using the GUI:

In order to use the GUI, you just have to open your browser and go to http://localhost:5001

You should see something like this:

Here you can use the form to search for companies' symbols. Select one from the dropdown list and then specify if you want to predict for a given number of days or a date range. Specify the number of days or the date range and click on the button. A panel will be added to the right, displaying the stock price along with the predicted data.