https://github.com/dioptx/volumepredictionassignment

Quick assignment for skill showcase
https://github.com/dioptx/volumepredictionassignment

Last synced: 9 months ago
JSON representation

Quick assignment for skill showcase

• Host: GitHub
• URL: https://github.com/dioptx/volumepredictionassignment
• Owner: dioptx
• Created: 2019-05-31T11:58:01.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2019-10-02T20:02:41.000Z (over 6 years ago)
• Last Synced: 2025-09-01T18:43:13.861Z (9 months ago)
• Language: Jupyter Notebook
• Size: 2.26 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Volume Stream Prediction - Tech: LSTM & SARIMA

---

### Code Coverage: ~85%

### Setup:

### Step 1:

#####  | Dependencies:

   1. Create a pip / conda Python 3.7 environment.

   2. Navigate into the project tree.

   3. Run the following command:

    

    # Unix

    sudo pip install -r requirements.txt

    # Windows (Elevated)

    pip install -r requirements.txt

----

### Step 2:

#####  | routes.csv import:

In order for all the tests to be run successfully

 place the routes.csv file that was given in the project description like this:

  

 - module

    - data

      - raw

        - routes.csv

        

----

The prediction functionality can be shown in the respective notebooks for SARIMA and LSTM models,

and follows the same mentality like the one bellow:

    from src.modeling import lstmModel

    from src.processing import dataProc

    

    # Aggregated 1 hour Dataset fetch

    DATASET_PATH = module_path + "/notebook/dt_agg1hour.h5"

    dataset = pd.read_pickle(

        DATASET_PATH)

    

    dataset = dataProc.create_features(dataset= dataset)

    # Model creation

    lstmodel = lstmModel(perform_scale=True)

    

    # Model training

    lstmodel.train(dataset[start_date:end_date], evaluate=False)

    

    # Model usage

    y_pred, y = lstmodel.generate_prediction(input_data= dataset[end_date:pd.Timestamp(end_date)+pd.Timedelta(hours=48)])