A Flask application to upload datasets, choose classifier, train classifier and predict unforeseen results. This app uses Celery to train data in background without making difficulties for main performance of the app.
Prior to installing the application doing some steps are critical. Using Anaconda is recommended and installation process is preformed with that assumption.
In order to use this code you need to install RabbitMQ broker. The installation is simple and you can find useful information about it here.
If you are in Ubuntu 20.04 you can simply install it by:
sudo apt-get install rabbitmq-server
You can find and download the best version of Anaconda for your OS here. Make sure you install Anaconda3 so that you can use Python3.
After you have installed RabbitMQ and Anaconda3, create an Anaconda environment is the first step:
conda create --name arminenv python=3.7
following the installation, activating environment is the next thing to do:
conda activate arminenv
Note: arminenv is the name of your environment. If you use a different name while creating it, use that name to activate the environment.
Download this repository and run the code below while you are inside the repo-folder:
pip install -r requirements.txt
Because this application uses celery to simulate doing a task on another server, you two separate terminals to run.
Inside repo folder, open a terminal and run the code below:
celery -A tasks worker --loglevel=INFO
This will initiate the Celery worker and activates background processing.
Note: Keep this terminal runing and do not close it while your app is the run.
Inside repo folder, open another terminal and run the code below:
python app.py
Next, you can use the app inside your browser at:
http://localhost:5000/
This application is a simple user-friendly web-app to train a database and classify unforeseen inputs.
Prior to any further steps, your database must follow a certain order.
- Your database should be in
CSVformat - Each row represents an input data
- The last column of each row is the Lablel for that input data
- Labels must be in Numbers. Any string label is not acceptable
There is Iris_flowers.csv inside Example Dataset that you can use as an example database. You can find more information about Iris_flowers dataset here.
If you run the application for the first time, you can see there is an empty table that should store your training history.
In order to upload a new database hit the Upload Dataset button.
It will lead you to another page where you can choose your desired data base. After choosing your data base, hit Submit button to start uploading process.
When uploading is done, you will return to Dashboard but a new row is added to the table. Important information such as File_ID, Upload Date, Trained Model and Status can be seen.
In order to train a database hit Train it! button.
On Train Dataset page, there is 3 classifier that can be chosen for your data. Select one and set parameters as you wish then hit Train it! button.
Training process happens on Celery therefore, the app leads you to Dashboard immediately. You will see gray badge in front of your File_ID saying Training... which means training hasn't finished yet.
If you refresh your Dashboard after some seconds, a Predict button will appear (Retrain button let you train a different classifier with new settings on your database).
Using Predict button, Predict webpage appears.
Based on your database and number of its attributes, you have to enter an input value and hit Submit button.
Your prediction will be shown next:
