s1-b7-sales prediction using ml

sales_prediction/.gitignore

@@ -0,0 +1,2 @@
+\venv
+requirements.txt

sales_prediction/CONTRIBUTORS.md

@@ -0,0 +1,5 @@
+
+- [AMIDELA HIMAJA](https://github.com/HIMAJA0710)
+- [DIDDI PRANAY](https://github.com/Pranay-Diddi)
+- [BAKTHULA LALITH PRABHASITH](https://github.com/JohnPrabhasith)
+- [A.V.GNANESWAR REDDY](https://github.com/gnaneswarr93)

sales_prediction/README.md

@@ -0,0 +1,54 @@
+# Sales Prediction Project
+
+This project focuses on predicting future sales based on historical data using machine learning techniques. It was created as part of a course/project work.
+
+## Introduction
+
+In retail and business management, predicting sales accurately is crucial for planning inventory, managing resources, and overall business strategy. This project aims to build a machine learning model and deploy it using Streamlit for interactive sales prediction.
+
+## Project Structure
+
+- `Sales Prediction.ipynb`: Jupyter notebook containing the entire project, including data preprocessing, model training, evaluation, and visualization.
+
+- `requirements.txt`: File listing the required Python packages for running the project.
+
+## Installation
+
+To run the project locally, follow these steps:
+
+1. Clone the repository:
+    ```sh
+    git clone https://github.com/HIMAJA0710/machine_learning/tree/s1-b7-sales_prediction/sales_prediction
+    ```
+
+2. Install the required dependencies:
+    ```sh
+    pip install -r requirements.txt
+    ```
+
+## Usage
+
+1. Open `Sales Prediction.ipynb` in Jupyter Notebook or JupyterLab.
+
+2. Execute the notebook cells to preprocess data, train the model, and evaluate it.
+
+3. Optionally, deploy the model using Streamlit:
+    ```sh
+    streamlit run app.py
+    ```
+
+4. Follow the instructions in the notebook or Streamlit app to interact with the sales prediction model.
+
+## Results
+
+The notebook and Streamlit app provide interactive visualizations and insights into:
+- Predicted vs Actual Sales
+- Model Performance Metrics (RMSE, MAE, R² score)
+- Feature Importance
+
+![alt text](<Screenshot 2024-06-21 202327[1].png>)
+
+
+![alt text](<Screenshot 2024-06-21 202246[1].png>)
+
+![alt text](<Screenshot 2024-06-21 202301[1].png>)

sales_prediction/app.py

@@ -0,0 +1,275 @@
+import streamlit as st
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sb
+from sklearn.model_selection import train_test_split
+from sklearn.ensemble import RandomForestRegressor
+from sklearn.metrics import mean_squared_error, r2_score
+import xgboost as xgb
+
+# Define functions for data processing and feature engineering
+def build_features(train, store):
+    store['StoreType'] = store['StoreType'].astype('category').cat.codes
+    store['Assortment'] = store['Assortment'].astype('category').cat.codes
+    train["StateHoliday"] = train["StateHoliday"].astype('category').cat.codes
+
+    merged = pd.merge(train, store, on='Store', how='left')
+    NaN_replace = 0
+    merged.fillna(NaN_replace, inplace=True)
+
+    merged['Year'] = merged.Date.dt.year
+    merged['Month'] = merged.Date.dt.month
+    merged['Day'] = merged.Date.dt.day
+    merged['Week'] = merged.Date.dt.isocalendar().week
+
+    merged['MonthsCompetitionOpen'] = \
+        12 * (merged['Year'] - merged['CompetitionOpenSinceYear']) + \
+        (merged['Month'] - merged['CompetitionOpenSinceMonth'])
+    merged.loc[merged['CompetitionOpenSinceYear'] == NaN_replace, 'MonthsCompetitionOpen'] = NaN_replace
+
+    merged['WeeksPromoOpen'] = \
+        12 * (merged['Year'] - merged['Promo2SinceYear']) + \
+        (merged['Date'].dt.isocalendar().week - merged['Promo2SinceWeek'])
+    merged.loc[merged['Promo2SinceYear'] == NaN_replace, 'WeeksPromoOpen'] = NaN_replace
+
+    toInt = [
+        'CompetitionOpenSinceMonth',
+        'CompetitionOpenSinceYear',
+        'Promo2SinceWeek',
+        'Promo2SinceYear',
+        'MonthsCompetitionOpen',
+        'WeeksPromoOpen'
+    ]
+    merged[toInt] = merged[toInt].astype(int)
+
+    return merged
+
+# Load data
+@st.cache_data
+def load_data():
+    types = {'StateHoliday': np.dtype(str)}
+    train = pd.read_csv("train_v2.csv", parse_dates=[2], nrows=66901, dtype=types)
+    store = pd.read_csv("store.csv")
+    return train, store
+
+train, store = load_data()
+
+# Data Cleaning
+train = train[train['Sales'] > 0]
+
+# Feature Engineering
+train['SalesPerCustomer'] = train['Sales'] / train['Customers']
+avg_store = train.groupby('Store')[['Sales', 'Customers', 'SalesPerCustomer']].mean()
+avg_store.rename(columns=lambda x: 'Avg' + x, inplace=True)
+store = pd.merge(avg_store.reset_index(), store, on='Store')
+
+med_store = train.groupby('Store')[['Sales', 'Customers', 'SalesPerCustomer']].median()
+med_store.rename(columns=lambda x: 'Med' + x, inplace=True)
+store = pd.merge(med_store.reset_index(), store, on='Store')
+
+features = build_features(train, store)
+
+# Sidebar for User Input
+st.sidebar.header('User Input Features')
+
+selected_store = st.sidebar.selectbox('Store', sorted(train['Store'].unique()))
+selected_day_of_week = st.sidebar.selectbox('Day of Week', sorted(train['DayOfWeek'].unique()))
+selected_competition_distance = st.sidebar.slider('Competition Distance (meters)', min_value=0, max_value=10000, step=100, value=1000)
+selected_promo = st.sidebar.selectbox('Promo', [0, 1])
+selected_promo2 = st.sidebar.selectbox('Promo2', [0, 1])
+selected_state_holiday = st.sidebar.selectbox('State Holiday', sorted(train['StateHoliday'].unique()))
+selected_store_type = st.sidebar.selectbox('Store Type', sorted(store['StoreType'].unique()))
+selected_assortment = st.sidebar.selectbox('Assortment', sorted(store['Assortment'].unique()))
+selected_date = st.sidebar.date_input('Select Date')
+
+# Filter data based on user input
+filtered_data = features[
+    (features['Store'] == selected_store) &
+    (features['DayOfWeek'] == selected_day_of_week)
+]
+
+# Display the filtered data
+st.write(f'Data for Store {selected_store} on Day {selected_day_of_week}')
+st.write(filtered_data)
+
+# Data Visualization
+def plot_scatter(data, x, y, hue=None):
+    plt.figure(figsize=(10, 6))
+    if hue:
+        sb.scatterplot(data=data, x=x, y=y, hue=hue)
+    else:
+        sb.scatterplot(data=data, x=x, y=y)
+    st.pyplot(plt)
+
+def plot_sales_distribution(data):
+    plt.figure(figsize=(10, 6))
+    sb.histplot(data['Sales'], kde=True)
+    plt.title('Sales Distribution')
+    plt.xlabel('Sales')
+    plt.ylabel('Frequency')
+    st.pyplot(plt)
+
+plot_scatter(filtered_data, 'Customers', 'Sales', 'StateHoliday')
+plot_sales_distribution(filtered_data)
+
+# Model Training and Evaluation
+X = [
+    'Store',
+    'Customers',
+    'CompetitionDistance',
+    'Promo',
+    'Promo2',
+    'StateHoliday',
+    'StoreType',
+    'Assortment',
+    'AvgSales',
+    'AvgCustomers',
+    'AvgSalesPerCustomer',
+    'MedSales',
+    'MedCustomers',
+    'MedSalesPerCustomer',
+    'DayOfWeek',
+    'Week',
+    'Day',
+    'Month',
+    'Year',
+    'CompetitionOpenSinceMonth',
+    'CompetitionOpenSinceYear',
+    'Promo2SinceWeek',
+    'Promo2SinceYear',
+]
+
+X_train, X_test, y_train, y_test = train_test_split(
+    features[X], features['Sales'], test_size=0.15, random_state=10)
+
+# Train RandomForest model
+randomForest = RandomForestRegressor(n_estimators=25, n_jobs=-1, verbose=1)
+randomForest.fit(X_train, y_train)
+
+# Display feature importance
+def plot_feature_importance(model, features):
+    importances = model.feature_importances_
+    indices = np.argsort(importances)
+    plt.figure(figsize=(12, 8))
+    plt.title('Feature Importances')
+    plt.barh(range(len(indices)), importances[indices], color='b', align='center')
+    plt.yticks(range(len(indices)), [features[i] for i in indices])
+    plt.xlabel('Relative Importance')
+    st.pyplot(plt)
+
+plot_feature_importance(randomForest, X)
+
+# Predict and display results
+y_hat = randomForest.predict(X_test)
+st.write('Random Forest Predictions')
+st.write(pd.DataFrame({'Actual': y_test, 'Predicted': y_hat}).head(10))
+
+# Model Performance Metrics for RandomForest
+rf_rmse = np.sqrt(mean_squared_error(y_test, y_hat))
+rf_r2 = r2_score(y_test, y_hat)
+st.write('Random Forest Model Performance')
+st.write(f'Root Mean Squared Error: {rf_rmse}')
+st.write(f'R-squared: {rf_r2}')
+
+# XGBoost model training
+xgboost_tree = xgb.XGBRegressor(
+    n_jobs=-1,
+    n_estimators=1000,
+    eta=0.1,
+    max_depth=2,
+    min_child_weight=2,
+    subsample=0.8,
+    colsample_bytree=0.8,
+    tree_method='exact',
+    reg_alpha=0.05,
+    silent=0,
+    random_state=1023
+)
+xgboost_tree.fit(X_train, np.log1p(y_train))
+
+# Display XGBoost feature importance
+def plot_xgb_feature_importance(model, features):
+    importance = model.feature_importances_
+    indices = np.argsort(importance)
+    plt.figure(figsize=(12, 8))
+    plt.title('XGBoost Feature Importances')
+    plt.barh(range(len(indices)), importance[indices], color='b', align='center')
+    plt.yticks(range(len(indices)), [features[i] for i in indices])
+    plt.xlabel('Relative Importance')
+    st.pyplot(plt)
+
+plot_xgb_feature_importance(xgboost_tree, X)
+
+# Predict and display results for XGBoost
+xgb_predictions = np.expm1(xgboost_tree.predict(X_test))
+st.write('XGBoost Predictions')
+st.write(pd.DataFrame({'Actual': y_test, 'Predicted': xgb_predictions}).head(10))
+
+# Model Performance Metrics for XGBoost
+xgb_rmse = np.sqrt(mean_squared_error(y_test, xgb_predictions))
+xgb_r2 = r2_score(y_test, xgb_predictions)
+st.write('XGBoost Model Performance')
+st.write(f'Root Mean Squared Error: {xgb_rmse}')
+st.write(f'R-squared: {xgb_r2}')
+
+# Final Prediction Based on User Input
+user_input = pd.DataFrame({
+    'Store': [selected_store],
+    'DayOfWeek': [selected_day_of_week],
+    'CompetitionDistance': [selected_competition_distance],
+    'Promo': [selected_promo],
+    'Promo2': [selected_promo2],
+    'StateHoliday': [selected_state_holiday],
+    'StoreType': [selected_store_type],
+    'Assortment': [selected_assortment],
+    'Year': [selected_date.year],
+    'Month': [selected_date.month],
+    'Day': [selected_date.day],
+    'Week': [pd.to_datetime(selected_date).isocalendar().week]
+})
+
+# Adding dummy 'Customers' column to user_input
+user_input['Customers'] = 0
+
+# Filling missing columns in user_input based on store data
+store_data = store[store['Store'] == selected_store]
+for col in ['AvgSales', 'AvgCustomers', 'AvgSalesPerCustomer', 'MedSales', 'MedCustomers', 'MedSalesPerCustomer']:
+    user_input[col] = store_data[col].values[0]
+
+# Ensure all necessary columns are present in user_input
+if 'CompetitionOpenSinceMonth' not in user_input.columns:
+    user_input['CompetitionOpenSinceMonth'] = store_data['CompetitionOpenSinceMonth'].values[0]
+if 'CompetitionOpenSinceYear' not in user_input.columns:
+    user_input['CompetitionOpenSinceYear'] = store_data['CompetitionOpenSinceYear'].values[0]
+if 'Promo2SinceWeek' not in user_input.columns:
+    user_input['Promo2SinceWeek'] = store_data['Promo2SinceWeek'].values[0]
+if 'Promo2SinceYear' not in user_input.columns:
+    user_input['Promo2SinceYear'] = store_data['Promo2SinceYear'].values[0]
+
+# Handle NaN values in user_input before type casting
+user_input.fillna(0, inplace=True)
+
+user_input['MonthsCompetitionOpen'] = 12 * (user_input['Year'] - user_input['CompetitionOpenSinceYear']) + (user_input['Month'] - user_input['CompetitionOpenSinceMonth'])
+user_input['WeeksPromoOpen'] = 12 * (user_input['Year'] - user_input['Promo2SinceYear']) + (user_input['Week'] - user_input['Promo2SinceWeek'])
+
+# Predict sales for the user input using RandomForest
+user_pred_rf = randomForest.predict(user_input[X])
+
+# Convert user_input to the format expected by XGBoost model
+user_input_xgb = user_input[X].copy()
+user_input_xgb['StateHoliday'] = user_input_xgb['StateHoliday'].astype('category').cat.codes
+
+# Predict with XGBoost
+user_pred_xgb = np.expm1(xgboost_tree.predict(user_input_xgb))
+
+st.write(f'Predicted Sales (Random Forest): ${user_pred_rf[0]:.2f}')
+st.write(f'Predicted Sales (XGBoost): ${user_pred_xgb[0]:.2f}')
+
+# Download filtered data button
+st.download_button(
+    label="Download Filtered Data",
+    data=filtered_data.to_csv(index=False).encode('utf-8'),
+    file_name='filtered_data.csv',
+    mime='text/csv',
+)