In this video tutorial, We explain how to create a Machine Learning Dashboard using Python dashboarding library "Streamlit". We train a random forest classifier from scikit-learn on a tiny dataset named wine dataset available from scikit-learn. After the training model, we evaluate various ML metrics (confusion matrix, classification report, feature importance, etc) and display them as charts on our dashboard. The dashboard also lets users try different feature values to see how the model performs for selected values. It even shows feature contributions as a bar chart.

Source Code: 

import streamlit as st

from sklearn.datasets import load_wine
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import confusion_matrix, accuracy_score, classification_report
from joblib import load

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import scikitplot as skplt

from lime import lime_tabular

## Load Data
wine = load_wine()
wine_df = pd.DataFrame(wine.data, columns=wine.feature_names)
wine_df["WineType"] = [wine.target_names[typ] for typ in wine.target]

X_train, X_test, Y_train, Y_test = train_test_split(wine.data, wine.target, train_size=0.8, random_state=123)

## Load Model
rf_classif = load("rf_classif.model")

Y_test_preds = rf_classif.predict(X_test)

## Dashboard
st.title("Wine Type :red[Prediction] :bar_chart: :chart_with_upwards_trend: :tea: :coffee:")
st.markdown("Predict Wine Type using Ingredients Values")

tab1, tab2, tab3 = st.tabs(["Data :clipboard:", "Global Performance :weight_lifter:", "Local Performance :bicyclist:"])

with tab1:
    st.header("Wine Dataset")
    st.write(wine_df)

with tab2:
    st.header("Confusion Matrix | Feature Importances")
    col1, col2 = st.columns(2)
    with col1:
        conf_mat_fig = plt.figure(figsize=(6,6))
        ax1 = conf_mat_fig.add_subplot(111)
        skplt.metrics.plot_confusion_matrix(Y_test, Y_test_preds, ax=ax1, normalize=True)
        st.pyplot(conf_mat_fig, use_container_width=True)

    with col2:
        feat_imp_fig = plt.figure(figsize=(6,6))
        ax1 = feat_imp_fig.add_subplot(111)
        skplt.estimators.plot_feature_importances(rf_classif, feature_names=wine.feature_names, ax=ax1, x_tick_rotation=90)
        st.pyplot(feat_imp_fig, use_container_width=True)

    st.divider()
    st.header("Classification Report")
    st.code(classification_report(Y_test, Y_test_preds))

with tab3:
    sliders = []
    col1, col2 = st.columns(2)
    with col1:
        for ingredient in wine.feature_names:
            ing_slider = st.slider(label=ingredient, min_value=float(wine_df[ingredient].min()), max_value=float(wine_df[ingredient].max()))
            sliders.append(ing_slider)

    with col2:
        col1, col2 = st.columns(2, gap="medium")
        
        prediction = rf_classif.predict([sliders])
        with col1:
            st.markdown("### Model Prediction : <strong style='color:tomato;'>{}</strong>".format(wine.target_names[prediction[0]]), unsafe_allow_html=True)

        probs = rf_classif.predict_proba([sliders])
        probability = probs[0][prediction[0]]

        with col2:
            st.metric(label="Model Confidence", value="{:.2f} %".format(probability*100), delta="{:.2f} %".format((probability-0.5)*100))

        explainer = lime_tabular.LimeTabularExplainer(X_train, mode="classification", class_names=wine.target_names, feature_names=wine.feature_names)
        explanation = explainer.explain_instance(np.array(sliders), rf_classif.predict_proba, num_features=len(wine.feature_names), top_labels=3)
        interpretation_fig = explanation.as_pyplot_figure(label=prediction[0])
        st.pyplot(interpretation_fig, use_container_width=True)

Dashboard Code: https://drive.google.com/file/d/1ZJ9TEdOnK1SQdyGY0ClCRkbw9WJjvML4/view?usp=sharing 

Subscribe: https://www.youtube.com/@CoderzColumn/featured 

#python #streamlit