import streamlit as st
import pandas as pd
import numpy as np
import plotly.express as px

from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA

st.title('Principal Component Analysis Demo App 🤖')
st.subheader('Credit: @BenjaminJack')
st.subheader('Raw Data')

iris = px.data.iris().drop('species_id',axis=1)
st.write(iris)

xvar = st.selectbox('Select x-axis',iris.columns[:-1])
yvar = st.selectbox('Select y-axis',iris.columns[:-1])

st.write(px.scatter(iris, x=xvar, y=yvar, color='species'))
iris_scaled = StandardScaler().fit_transform(iris.drop('species',axis=1))
iris_pca = PCA()

iris_transformed = iris_pca.fit_transform(iris_scaled)
col_names = [f'component{i+1}' for i in range(iris_transformed.shape[1])]
st.write(col_names)

iris_transformed_df = pd.DataFrame(iris_transformed, columns=col_names)
iris_transformed_df = pd.concat([iris_transformed_df, iris['species']],axis=1)

st.subheader('Transformed Data')

st.write(iris_transformed_df)

st.subheader('Explore principal components')
xvar = st.selectbox('Select x-axis:', iris_transformed_df.columns[:-1])
yvar = st.selectbox('Select y-axis:', iris_transformed_df.columns[:-1])

st.write(px.scatter(iris_transformed_df,x=xvar, y=yvar, color = 'species'))

st.subheader('Explore Loadings')

loadings = iris_pca.components_.T * np.sqrt(iris_pca.explained_variance_)

loadings_df = pd.DataFrame(loadings,columns=col_names)
loadings_df = pd.concat([loadings_df,
                          pd.Series(iris.columns[0:4],name='var')],
                          axis=1)

component =st.selectbox('Select Component:',loadings_df.columns[0:4])

bar_chart = px.bar(loadings_df[['var',component]].sort_values(component),
                   y='var',
                   x=component,
                   orientation='h',
                   range_x=[-1,1]
                   )
st.write(bar_chart)
st.write(loadings_df)

#Credits : benjaminjack