AutoML-Benchmark

Benchmark for some usual automated machine learning, such as: AutoSklearn, MLJAR, H2O, TPOT and AutoGluon. All visualized via a responsive Dash Ploty Web Application.

Requirements and Installation

The python version for this project is the 3.9, so make sure to have installed First of all install the python3.9 pip package and then the virtual environment for safety reason:

curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
python3.9 get-pip.py
sudo apt install default-jre
sudo apt install python3.9-venv

Then create a new virtual environment:

python3.9 -m venv my_venv

Access at it and activate it:

cd my_venv
source bin/activate

Clone my repository and install all dependencies with:

make install

Usage

To run the app execute the following line of code:

python3 start.py

Open your favourite browser and go to: http://127.0.0.1:8050/. Here you will be albe to interact with the application

Type of tests

There are five types of operations:

1. OpenML Benchmark: Here you have two options:

   1. You can insert a sequrnce of dataframe ID where each ID is followed by a comma. This command will run a benchmark on the inserted sequence
   2. Or you can choose the number of dataframes each for classification task and for regression task and the number of instances that a dataframe at least has. This command will start a benchmark using openml dataframes

2. Kaggle Benchmark: Here you can choose multiple kaggle's dataframes for running a benchmark on them

3. Test Benchmark: Here you can run a benchmark on a specific dataframe by insering the dataframe id and using a single algorithm ot all of them by selecting a options

4. Past Results OpenML: Here you can navigate between past OpenML benchmark by selecting a specific date, or you can comapre multiple OpenML Benchmarks that have the same dataframes but with different timelife

5. Past Results Kaggle: Here you can navigate between past Kaggle benchmark by selecting a specific date, or you can comapre multiple Kaggle Benchmarks that have the same dataframes but with different timelife

Actions available

In all operation these action are available:

• Analize the results of Classification Tasks and Regression Tasks by a Table visualization, Bar Charts visualization and Scatter Plot visualization
• See the Timelife of all algorithms
• Inspect the Pipeline of al algorithms