[PYTHON] Principal component analysis hands-on with PyCaret [normalization + visualization (plotly)] memo

Overview

• PCA was done with PyCaret. (Hyperparameters etc. are not optimized)
• PCA was performed with certain electric energy data.

Source code

• Own data formatting source code (dirty): 13 lines

Own data formatting code

import pandas as pd
def load_sampledata():
    data = pd.read_csv(r'pycaret_sample.csv',
            encoding='shift-jis',
            engine='python',
            index_col=[0],
            parse_dates=[0])

    data = data.resample('h').sum()
    data['hour'] = data.index.hour
    data['date'] = data.index.strftime('%Y-%m-%d (%a)')
    dataset = data.pivot(index='hour',columns='date',values='Electric energy')
    dataset = dataset.T.reset_index()
    return dataset

• PyCaret setup: 2 lines

setup

from pycaret.clustering import * 
clu = setup(data = load_sampledata(),normalize=True)

• You can check the setup contents

• PyCaret modeling + plotting: 2 lines

create_model&plot

# creating a model
hclust = create_model('hclust')
# plotting a model
plot_model(hclust,plot='cluster',feature='date')

• You can even visualize PCA in 2 lines.

Summary

• PyCaret is amazing.
• If you start writing with this, you will not be able to write without a wrapper.
• You can tune and write very quickly. It's amazing.