sn_ml_cloud / src /streamlit_app.py
DavidNgoue's picture
Update src/streamlit_app.py
de623e1 verified
Raw History Blame Contribute Delete
24.7 kB
import streamlit as st
import pandas as pd
import numpy as np
import pickle
from PIL import Image
import plotly.express as px
import plotly.graph_objects as go
from streamlit_lottie import st_lottie
import requests
from streamlit_option_menu import option_menu
# --------- UTILS ---------
@st.cache_resource
def load_model():
with open('src/best_regression_model.pkl', 'rb') as f:
data = pickle.load(f)
return data['model'], data['scaler_X'], data['scaler_y'], data.get('metrics', None)
def load_lottieurl(url: str):
r = requests.get(url)
if r.status_code != 200:
return None
return r.json()
# --------- ANIMATIONS ---------
credit_animation = load_lottieurl("https://assets2.lottiefiles.com/packages/lf20_4kx2q32n.json")
loading_animation = load_lottieurl("https://assets3.lottiefiles.com/packages/lf20_p8bfn5to.json")
about_animation = load_lottieurl("https://assets2.lottiefiles.com/packages/lf20_0yfsb3a1.json")
# --------- PAGE CONFIG ---------
st.set_page_config(
page_title="Credit Card Expenditure Predictor",
page_icon="💳",
layout="wide",
initial_sidebar_state="expanded"
)
# --------- CSS ---------
st.markdown("""
<style>
@import url('https://fonts.googleapis.com/css2?family=Poppins:wght@400;600&display=swap');
* { font-family: 'Poppins', sans-serif; }
.main { background: linear-gradient(135deg, #f5f7fa 0%, #e4e8eb 100%); }
.stButton>button {
background: linear-gradient(45deg, #4b79a1, #283e51);
color: white;
border-radius: 25px;
padding: 10px 25px;
border: none;
box-shadow: 0 4px 15px rgba(75, 121, 161, 0.3);
transition: all 0.3s ease;
font-weight: 600;
font-size: 1.1em;
}
.stButton>button:hover {
transform: translateY(-2px) scale(1.04);
box-shadow: 0 6px 20px rgba(75, 121, 161, 0.4);
background: linear-gradient(45deg, #283e51, #4b79a1);
}
.sidebar .sidebar-content {
background: linear-gradient(180deg, #ffffff 0%, #f8f9fa 100%);
box-shadow: 2px 0 10px rgba(0,0,0,0.1);
}
.stSelectbox, .stNumberInput { border-radius: 10px; }
.stProgress > div > div { background-color: #4b79a1; }
.stMarkdown { color: #2c3e50; }
.stAlert { border-radius: 10px; }
.section-card {
background: white;
padding: 2rem;
border-radius: 18px;
box-shadow: 0 8px 32px 0 rgba(76, 110, 245, 0.10);
margin-bottom: 2rem;
transition: box-shadow 0.3s;
}
.section-card:hover {
box-shadow: 0 16px 40px 0 rgba(76, 110, 245, 0.18);
}
.section-title {
color: #4b79a1;
font-size: 2.2em;
font-weight: 700;
margin-bottom: 0.5em;
text-shadow: 1px 1px 2px #e4e8eb;
}
.badge {
display: inline-block;
background: linear-gradient(90deg, #4b79a1 0%, #283e51 100%);
color: white;
border-radius: 12px;
padding: 0.3em 1em;
font-size: 1em;
font-weight: 600;
margin: 0.2em 0.3em;
box-shadow: 0 2px 8px rgba(76, 110, 245, 0.10);
}
.about-avatar {
border-radius: 50%;
border: 4px solid #4b79a1;
box-shadow: 0 4px 16px rgba(76, 110, 245, 0.15);
margin-bottom: 1em;
}
.about-contact-btn {
background: linear-gradient(90deg, #4b79a1 0%, #283e51 100%);
color: white;
border-radius: 20px;
padding: 0.5em 1.5em;
border: none;
font-weight: 600;
margin: 0.5em 0.5em 0.5em 0;
font-size: 1.1em;
box-shadow: 0 2px 8px rgba(76, 110, 245, 0.10);
transition: background 0.2s;
}
.about-contact-btn:hover {
background: linear-gradient(90deg, #283e51 0%, #4b79a1 100%);
color: #fff;
}
.card-fade {
opacity: 0;
transform: translateY(30px);
animation: fadeInUp 0.8s forwards;
animation-delay: 0.2s;
}
@keyframes fadeInUp {
to {
opacity: 1;
transform: none;
}
}
.section-sep {
border: none;
border-top: 2px solid #e4e8eb;
margin: 2.5em 0 2em 0;
width: 80%;
}
.section-title-visual {
font-size: 2em;
color: #283e51;
font-weight: 700;
margin-bottom: 0.7em;
letter-spacing: 1px;
text-shadow: 0 2px 8px #e4e8eb;
}
.visual-card {
background: linear-gradient(120deg, #fafdff 0%, #f5f7fa 100%);
border-radius: 18px;
box-shadow: 0 4px 24px 0 rgba(76, 110, 245, 0.10);
padding: 1.5em 2em 1.5em 2em;
margin-bottom: 2.5em;
transition: box-shadow 0.3s, transform 0.3s;
}
.visual-card:hover {
box-shadow: 0 12px 32px 0 rgba(76, 110, 245, 0.18);
transform: translateY(-4px) scale(1.01);
}
.metric-card {
background: linear-gradient(90deg, #4b79a1 0%, #283e51 100%);
color: white;
border-radius: 18px;
box-shadow: 0 4px 24px 0 rgba(76, 110, 245, 0.10);
padding: 2em 2em 1.5em 2em;
margin-bottom: 2.5em;
text-align: center;
position: relative;
overflow: hidden;
}
.metric-card h3 {
font-size: 1.7em;
margin-bottom: 0.7em;
color: #fff;
letter-spacing: 1px;
}
.metric-list {
list-style: none;
padding: 0;
margin: 0 auto;
font-size: 1.15em;
}
.metric-list li {
margin: 0.7em 0;
padding: 0.5em 0;
border-bottom: 1px solid #ffffff22;
}
.metric-list li:last-child {
border-bottom: none;
}
</style>
""", unsafe_allow_html=True)
# --------- SIDEBAR ---------
with st.sidebar:
st_lottie(credit_animation, height=120, key="sidebar_animation")
selected = option_menu(
menu_title="Navigation",
options=["Accueil", "Prédiction", "Analyse", "À propos"],
icons=['house', 'credit-card', 'bar-chart', 'info-circle'],
menu_icon="cast",
default_index=0,
styles={
"container": {"padding": "0!important", "background-color": "#ffffff"},
"icon": {"color": "#4b79a1", "font-size": "20px"},
"nav-link": {
"font-size": "16px",
"text-align": "left",
"margin": "0px",
"padding": "10px",
"--hover-color": "#4b79a1",
},
"nav-link-selected": {"background-color": "#4b79a1"},
}
)
# --------- ACCUEIL ---------
if selected == "Accueil":
st.markdown("""
<div style='
text-align: center;
padding: 2.5rem 1rem 2rem 1rem;
background: linear-gradient(135deg, #f8fafc 0%, #e4e8eb 100%);
border-radius: 22px;
margin-bottom: 36px;
box-shadow: 0 6px 32px 0 rgba(76, 110, 245, 0.08);
'>
<h1 class='section-title' style='font-size:2.8em; margin-bottom:0.2em;'>💳 Credit Card Expenditure Predictor</h1>
<p style='color: #2c3e50; font-size: 1.35em; font-weight: 400; margin-bottom:0.8em;'>
<i>Prédisez les dépenses annuelles de vos clients grâce à l'IA, pour une gestion financière plus intelligente et personnalisée.</i>
</p>
<hr style='border: none; border-top: 1.5px solid #e4e8eb; width: 60%; margin: 1.5em auto 1.5em auto;'/>
<p style='color: #4b79a1; font-size: 1.1em; max-width: 700px; margin: auto;'>
Cette application met la puissance du machine learning au service de la finance :
<b>analysez, prédisez et optimisez</b> les dépenses de carte de crédit de vos clients en quelques clics.<br>
<span style='color:#283e51;'>Pensée pour les professionnels, accessible à tous.</span>
</p>
</div>
""", unsafe_allow_html=True)
col1, col2 = st.columns([1.5, 1])
with col1:
st.markdown("""
<div class='section-card' style='margin-bottom:1.5em;'>
<h3 style='color:#4b79a1; font-size:1.3em;'>🎯 Mission</h3>
<p style='font-size:1.08em;'>
Offrir un outil prédictif fiable et intuitif pour anticiper les dépenses annuelles des clients,
en s'appuyant sur leurs caractéristiques financières et personnelles.
</p>
</div>
<div class='section-card'>
<h3 style='color:#4b79a1; font-size:1.2em;'>🔬 Technologies</h3>
<span class='badge'>Random Forest</span>
<span class='badge'>XGBoost</span>
<span class='badge'>SVR</span>
<span class='badge'>GridSearchCV</span>
<span class='badge'>Scikit-learn</span>
<span class='badge'>Streamlit</span>
<span class='badge'>Plotly</span>
</div>
""", unsafe_allow_html=True)
with col2:
st_lottie(credit_animation, height=220, key="main_animation")
st.markdown("""
<div style='margin-top: 18px; background: linear-gradient(135deg, #4b79a1 0%, #283e51 100%); color: white; padding: 18px; border-radius: 12px; text-align: center; box-shadow: 0 2px 10px rgba(76,110,245,0.10);'>
<h2 style='margin:0;'>+2000</h2>
<p style='margin:0;'>Clients analysés</p>
</div>
""", unsafe_allow_html=True)
# --------- PREDICTION ---------
elif selected == "Prédiction":
st.markdown("""
<div style='
background: linear-gradient(120deg, #f8fafc 0%, #e4e8eb 100%);
padding: 2.2rem 1rem 2rem 1rem;
border-radius: 22px;
color: #283e51;
margin-bottom: 2.2rem;
text-align: center;
box-shadow: 0 4px 15px rgba(0,0,0,0.07);
'>
<h1 class='section-title' style='color:#4b79a1;'>🔮 Prédiction de Dépense</h1>
<p style='font-size:1.15em;'>Remplissez le formulaire ci-dessous pour estimer la dépense annuelle d'un client</p>
</div>
""", unsafe_allow_html=True)
with st.form("prediction_form"):
st.markdown("<div class='section-card' style='background:#fafdff;'>", unsafe_allow_html=True)
col1, col2 = st.columns(2)
with col1:
st.markdown("<h4 style='color:#4b79a1; margin-bottom:0.5em;'>Informations personnelles</h4>", unsafe_allow_html=True)
age = st.number_input("Âge", min_value=18, max_value=100, value=35)
owner = st.selectbox("Propriétaire d'une maison", ["Non", "Oui"])
selfemp = st.selectbox("Travailleur indépendant", ["Non", "Oui"])
dependents = st.number_input("Nombre de personnes à charge", min_value=0, max_value=10, value=0)
with col2:
st.markdown("<h4 style='color:#4b79a1; margin-bottom:0.5em;'>Informations financières</h4>", unsafe_allow_html=True)
income = st.number_input("Revenu annuel ($)", min_value=0, max_value=500000, value=50000)
share = st.slider("Part de revenu allouée à la carte (%)", min_value=0, max_value=100, value=10)
reports = st.number_input("Nombre de rapports de crédit", min_value=0, max_value=20, value=2)
months = st.number_input("Ancienneté (mois)", min_value=0, max_value=240, value=12)
majorcards = st.number_input("Nombre de cartes principales", min_value=0, max_value=5, value=1)
active = st.number_input("Nombre de comptes actifs", min_value=0, max_value=10, value=2)
st.markdown("<hr style='margin:1.5em 0;'/>", unsafe_allow_html=True)
real_expenditure = st.number_input("Valeur réelle de la dépense (optionnel)", min_value=0, max_value=100000, value=0)
submit_button = st.form_submit_button("💡 Prédire la dépense", use_container_width=True)
st.markdown("</div>", unsafe_allow_html=True)
if submit_button:
with st.spinner("Prédiction en cours..."):
st_lottie(loading_animation, height=100, key="loading")
# Préparation des données
input_df = pd.DataFrame({
'income': [income],
'share': [share],
'age': [age],
'owner_No': [1 if owner == "Non" else 0],
'owner_Yes': [1 if owner == "Oui" else 0],
'selfemp_No': [1 if selfemp == "Non" else 0],
'selfemp_Yes': [1 if selfemp == "Oui" else 0],
'reports': [reports],
'dependents': [dependents],
'months': [months],
'majorcards': [majorcards],
'active': [active]
})
# Charger modèle et scalers
model, scaler_X, scaler_y, metrics = load_model()
# Adapter les colonnes à l'ordre attendu
X_cols = scaler_X.feature_names_in_
input_df = input_df.reindex(columns=X_cols, fill_value=0)
# Normaliser
X_scaled = scaler_X.transform(input_df)
y_pred_scaled = model.predict(X_scaled)
y_pred = scaler_y.inverse_transform(y_pred_scaled.reshape(-1, 1)).ravel()[0]
# Affichage
st.markdown("""
<div class='section-card' style='margin-top:2rem;'>
<h2 class='section-title' style='text-align:center;'>Résultat de la Prédiction</h2>
""", unsafe_allow_html=True)
st.metric("Dépense prédite ($)", f"{y_pred:,.2f}")
if real_expenditure > 0:
st.metric("Valeur réelle ($)", f"{real_expenditure:,.2f}", delta=f"{y_pred-real_expenditure:,.2f}")
fig = go.Figure()
fig.add_trace(go.Bar(
x=["Prédiction", "Réel"],
y=[y_pred, real_expenditure],
marker_color=["#4b79a1", "#283e51"]
))
fig.update_layout(title="Comparaison Prédiction vs Réel", yaxis_title="Dépense ($)")
st.plotly_chart(fig, use_container_width=True)
st.markdown("</div>", unsafe_allow_html=True)
# --------- ANALYSE ---------
elif selected == "Analyse":
st.markdown("""
<div style='background: linear-gradient(120deg, #283e51 0%, #4b79a1 100%);
padding: 2.5rem 1rem 2rem 1rem;
border-radius: 22px;
color: white;
margin-bottom: 2.5rem;
text-align: center;
box-shadow: 0 4px 15px rgba(0,0,0,0.10);'>
<h1 class='section-title'>📊 Tableau de Bord Analytique</h1>
<p style='font-size:1.2em; color:#e4e8eb;'>Explorez la performance du modèle et les tendances clés du dataset.</p>
</div>
""", unsafe_allow_html=True)
# Charger le dataset et le modèle
df = pd.read_csv('AER_credit_card_data.csv')
X = df.drop(['expenditure', 'card'], axis=1)
y = df['expenditure']
X = pd.get_dummies(X, columns=['owner', 'selfemp'])
model, scaler_X, scaler_y, metrics = load_model()
X = X.reindex(columns=scaler_X.feature_names_in_, fill_value=0)
X_scaled = scaler_X.transform(X)
y_pred_scaled = model.predict(X_scaled)
y_pred = scaler_y.inverse_transform(y_pred_scaled.reshape(-1, 1)).ravel()
from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
mse = mean_squared_error(y, y_pred)
rmse = np.sqrt(mse)
mae = mean_absolute_error(y, y_pred)
r2 = r2_score(y, y_pred)
# --- Bloc Performance du Modèle ---
if metrics:
# On affiche uniquement les vraies métriques du best model
st.markdown(f"""
<div class="metric-card card-fade">
<h3>✨ Performance du Meilleur Modèle</h3>
<ul class="metric-list">
<li><b>RMSE (test)</b> : {metrics.get('rmse', 'N/A'):.2f}</li>
<li><b>MAE (test)</b> : {metrics.get('mae', 'N/A'):.2f}</li>
<li><b>R² (test)</b> : {metrics.get('r2', 'N/A'):.3f}</li>
{f"<li><b>Score CV</b> : {metrics['cv_score']:.3f}</li>" if 'cv_score' in metrics else ""}
</ul>
</div>
<hr class="section-sep"/>
<div class="section-title-visual">Analyse Visuelle</div>
""", unsafe_allow_html=True)
else:
# fallback si jamais metrics n'est pas dispo
st.markdown("""
<div class="metric-card card-fade">
<h3>✨ Performance du Modèle</h3>
<p style="color:#fff;">(Métriques calculées sur tout le dataset, à titre indicatif)</p>
<ul class="metric-list">
<li><b>RMSE</b> : {:.2f}</li>
<li><b>MAE</b> : {:.2f}</li>
<li><b>R²</b> : {:.3f}</li>
</ul>
</div>
<hr class="section-sep"/>
<div class="section-title-visual">Analyse Visuelle</div>
""".format(rmse, mae, r2), unsafe_allow_html=True)
# --- 1. Scatter plot Prédiction vs Réel ---
st.markdown("""
<div class="visual-card card-fade">
<h4 style='color:#4b79a1; margin-bottom:0.3em;'>1. Prédictions vs Valeurs réelles</h4>
<p style='color:#444; margin-bottom:1.2em;'>Chaque point représente un client. Plus les points sont proches de la diagonale, plus la prédiction est précise.</p>
""", unsafe_allow_html=True)
fig1 = px.scatter(
x=y, y=y_pred,
labels={'x': 'Valeur réelle', 'y': 'Prédiction'},
color_discrete_sequence=["#4b79a1"],
title=None
)
fig1.add_shape(
type="line",
x0=y.min(), y0=y.min(),
x1=y.max(), y1=y.max(),
line=dict(color="#e74c3c", dash="dash")
)
fig1.update_layout(showlegend=False, height=350, margin=dict(l=20, r=20, t=30, b=20))
st.plotly_chart(fig1, use_container_width=True)
st.markdown("</div>", unsafe_allow_html=True)
# --- 2. Distribution des Dépenses Réelles ---
st.markdown("""
<div class="visual-card card-fade">
<h4 style='color:#4b79a1; margin-bottom:0.3em;'>2. Distribution des dépenses réelles</h4>
<p style='color:#444; margin-bottom:1.2em;'>Visualisation de la répartition des dépenses annuelles des clients.</p>
""", unsafe_allow_html=True)
fig2 = px.histogram(df, x="expenditure", nbins=40, color_discrete_sequence=["#283e51"])
fig2.update_layout(
xaxis_title="Dépense annuelle ($)",
yaxis_title="Nombre de clients",
height=300,
margin=dict(l=20, r=20, t=30, b=20)
)
st.plotly_chart(fig2, use_container_width=True)
st.markdown("</div>", unsafe_allow_html=True)
# --- 3. Importance des variables ---
st.markdown("""
<div class="visual-card card-fade">
<h4 style='color:#4b79a1; margin-bottom:0.3em;'>3. Importance des variables</h4>
<p style='color:#444; margin-bottom:1.2em;'>Les variables les plus influentes dans la prédiction selon le modèle.</p>
""", unsafe_allow_html=True)
if hasattr(model, "feature_importances_"):
importances = model.feature_importances_
features = X.columns
imp_df = pd.DataFrame({"Variable": features, "Importance": importances})
imp_df = imp_df.sort_values("Importance", ascending=True)
fig3 = px.bar(
imp_df,
x="Importance", y="Variable",
orientation="h",
color="Importance",
color_continuous_scale="blues",
height=350
)
fig3.update_layout(margin=dict(l=20, r=20, t=30, b=20), coloraxis_showscale=False)
st.plotly_chart(fig3, use_container_width=True)
else:
st.info("L'importance des variables n'est pas disponible pour ce modèle.")
st.markdown("</div>", unsafe_allow_html=True)
# --- 4. Dépense moyenne par statut de propriétaire ---
st.markdown("""
<div class="visual-card card-fade">
<h4 style='color:#4b79a1; margin-bottom:0.3em;'>4. Dépense moyenne selon le statut de propriétaire</h4>
<p style='color:#444; margin-bottom:1.2em;'>Comparaison des dépenses annuelles entre propriétaires et non-propriétaires.</p>
""", unsafe_allow_html=True)
fig4 = px.box(
df, x="owner", y="expenditure",
color="owner",
color_discrete_sequence=["#4b79a1", "#283e51"],
points="all",
height=320
)
fig4.update_layout(
xaxis_title="Statut de propriétaire",
yaxis_title="Dépense annuelle ($)",
showlegend=False,
margin=dict(l=20, r=20, t=30, b=20)
)
st.plotly_chart(fig4, use_container_width=True)
st.markdown("</div>", unsafe_allow_html=True)
# --- 5. Matrice de corrélation ---
st.markdown("""
<div class="visual-card card-fade">
<h4 style='color:#4b79a1; margin-bottom:0.3em;'>5. Corrélation entre variables</h4>
<p style='color:#444; margin-bottom:1.2em;'>Les relations linéaires entre les principales variables du dataset.</p>
""", unsafe_allow_html=True)
corr = df.select_dtypes(include=[np.number]).corr()
fig5 = px.imshow(
corr,
text_auto=True,
color_continuous_scale="blues",
aspect="auto",
height=400
)
fig5.update_layout(margin=dict(l=20, r=20, t=30, b=20))
st.plotly_chart(fig5, use_container_width=True)
st.markdown("</div>", unsafe_allow_html=True)
# --------- A PROPOS ---------
elif selected == "À propos":
st.markdown("""
<div style='background: linear-gradient(120deg, #4b79a1 0%, #283e51 100%); padding: 2rem; border-radius: 18px; color: white; margin-bottom: 2rem; text-align: center; box-shadow: 0 4px 15px rgba(0,0,0,0.1);'>
<h1 class='section-title' style='color:white;'>À propos</h1>
<p>Découvrez le créateur, le projet et les technologies utilisées</p>
</div>
""", unsafe_allow_html=True)
col1, col2 = st.columns([1, 2])
with col1:
st_lottie(about_animation, height=220, key="about_animation")
st.image(
"https://avatars.githubusercontent.com/u/TheBeyonder237",
width=180,
caption="Ngoue David",
output_format="auto",
use_column_width=False,
channels="RGB"
)
st.markdown("""
<div style='text-align:center; margin-top:1em;'>
<button class='about-contact-btn' onclick="window.open('mailto:ngouedavidrogeryannick@gmail.com')">📧 Email</button>
<button class='about-contact-btn' onclick="window.open('https://github.com/TheBeyonder237')">🌐 GitHub</button>
</div>
""", unsafe_allow_html=True)
with col2:
st.markdown("""
<div class='section-card'>
<h2 class='section-title'>Qui suis-je ?</h2>
<p>
Je suis un passionné de l'intelligence artificielle et de la donnée.<br>
Actuellement en Master 2 en IA et Big Data, je travaille sur des solutions innovantes dans le domaine de l'Intelligence Artificielle appliquée à la finance et à la santé.
</p>
<h3 style='color:#4b79a1;'>Compétences</h3>
<span class='badge'>Python</span>
<span class='badge'>Machine Learning</span>
<span class='badge'>Deep Learning</span>
<span class='badge'>NLP</span>
<span class='badge'>Data Science</span>
<span class='badge'>Cloud Computing</span>
<span class='badge'>Streamlit</span>
<span class='badge'>Scikit-learn</span>
<span class='badge'>XGBoost</span>
<span class='badge'>Pandas</span>
<span class='badge'>Plotly</span>
<span class='badge'>SQL</span>
<h3 style='color:#4b79a1; margin-top:1.5em;'>Projets Récents</h3>
<ul>
<li><b>💳 Credit Card Expenditure Predictor</b> : Application de prédiction de dépenses de carte de crédit.</li>
<li><b>🫀 HeartGuard AI</b> : Prédiction de risques cardiaques par IA.</li>
<li><b>🔊 Multi-IA</b> : Plateforme multi-modèles pour la génération de texte, synthèse vocale et traduction.</li>
</ul>
</div>
""", unsafe_allow_html=True)
st.markdown("""
<div style='text-align: center; color: #666; padding: 20px;'>
Développé avec ❤️ par Ngoue David
</div>
""", unsafe_allow_html=True)