from flask import Flask, render_template
import pandas as pd
from scipy.cluster.hierarchy import linkage, fcluster
import seaborn as sns
from sklearn.preprocessing import LabelEncoder
import io
import base64

app = Flask(__name__)

# Load data from student-mat.csv
data = pd.read_csv('student-mat.csv')

def generate_clusters_image():
    # Selecting columns of interest for clustering
    features = data[['Pstatus', 'guardian', 'internet', 'romantic', 'famrel', 'freetime', 'goout', 'Dalc', 'Walc', 'health', 'absences']]

    # Convert categorical variables to numerical using LabelEncoder
    label_encoder = LabelEncoder()
    features_categorical = features.select_dtypes(include='object')
    features[features_categorical.columns] = features_categorical.apply(label_encoder.fit_transform)

    # Calculate linkage matrix
    linkage_matrix = linkage(features, method='ward')

    # Cut the tree to form clusters
    data['cluster'] = fcluster(linkage_matrix, 3, criterion='maxclust')  # You can adjust the number of clusters

    # Drop the 'cluster' column to avoid confusion in sns.clustermap
    features_for_heatmap = features.copy()  # Make a copy to avoid modifying the original DataFrame
    features_for_heatmap['cluster'] = data['cluster']
    features_for_heatmap = features_for_heatmap.drop('cluster', axis=1)

    # Visualize clusters
    sns.set(style='white')
    g = sns.clustermap(features_for_heatmap, col_cluster=False, figsize=(24, 16), cmap="viridis")

    # Save the image to a buffer
    img_buffer = io.BytesIO()
    g.savefig(img_buffer, format='png')
    img_buffer.seek(0)

    # Convert the image to a base64 string
    img_base64 = base64.b64encode(img_buffer.read()).decode()

    return img_base64

@app.route('/')
def index():
    clusters_image = generate_clusters_image()
    return render_template('index.html', clusters_image=clusters_image)

if __name__ == '__main__':
    app.run(port=5000)