Spaces:
Sleeping
Sleeping
import streamlit as st | |
import pandas as pd | |
import numpy as np | |
import joblib | |
from sklearn.preprocessing import MinMaxScaler | |
import plotly.graph_objects as go | |
from keras.models import load_model | |
from datetime import datetime, timedelta | |
# Load the trained model | |
model = joblib.load('./lstm_model.pkl') | |
# Function to prepare the data | |
def prepare_data(df, time_steps=60): | |
data = df['quantity'].values.reshape(-1, 1) | |
scaler = MinMaxScaler(feature_range=(0, 1)) | |
scaled_data = scaler.fit_transform(data) | |
x_test = [] | |
for i in range(time_steps, len(scaled_data)): | |
x_test.append(scaled_data[i - time_steps:i, 0]) | |
x_test = np.array(x_test) | |
x_test = np.reshape(x_test, (x_test.shape[0], x_test.shape[1], 1)) | |
return x_test, scaler | |
# Function to forecast the next 60 days | |
def forecast(model, x_test, scaler, time_steps=60, future=60): | |
forecast_data = x_test[-1] # Use the last sequence of the test set for forecasting | |
forecast_predictions = [] | |
for _ in range(future): | |
prediction = model.predict(forecast_data.reshape(1, time_steps, 1)) | |
forecast_predictions.append(prediction[0, 0]) | |
forecast_data = np.append(forecast_data[1:], prediction[0, 0]).reshape(-1, 1) | |
forecast_predictions = np.array(forecast_predictions).reshape(-1, 1) | |
forecast_predictions = scaler.inverse_transform(forecast_predictions) | |
return forecast_predictions | |
# Streamlit UI | |
st.title('Product Sales Forecasting') | |
uploaded_file = st.file_uploader("Choose a file") | |
if uploaded_file is not None: | |
df = pd.read_csv(uploaded_file, parse_dates=['date']) | |
st.write(df.tail()) # Display the tail of the dataframe | |
family = st.selectbox("Select a family", df['family'].unique()) | |
if st.button('Predict'): | |
df_family = df[df['family'] == family] | |
# Ensure df_family is not empty | |
if df_family.empty: | |
st.write("No data available for the selected family.") | |
else: | |
# Prepare data | |
x_test, scaler = prepare_data(df_family) | |
# Forecast | |
forecast_predictions = forecast(model, x_test, scaler) | |
# Prepare forecast dataframe | |
last_date = df_family['date'].max() | |
forecast_dates = [last_date + timedelta(days=i) for i in range(1, 61)] | |
forecast_df = pd.DataFrame({'date': forecast_dates, 'forecasted_quantity': forecast_predictions.flatten()}) | |
# Plot using Plotly with green line | |
fig = go.Figure() | |
fig.add_trace(go.Scatter(x=forecast_df['date'], y=forecast_df['forecasted_quantity'], mode='lines', name='Forecasted Quantity', line=dict(color='green'))) | |
fig.update_layout(title=f'Sales Forecast for {family}', xaxis_title='Date', yaxis_title='Quantity Sold') | |
st.plotly_chart(fig) | |
st.write(forecast_df) | |