app.py

import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import torch
from chronos import ChronosPipeline

class TimeSeriesForecaster:
    def __init__(self, model_name="amazon/chronos-t5-small"):
        self.pipeline = ChronosPipeline.from_pretrained(
            model_name,
            device_map="cuda" if torch.cuda.is_available() else "cpu",
            torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32
        )
        self.original_series = None
        self.context = None

    def preprocess_data(self, df, date_column, value_column, context_length=100, prediction_length=30):
        """
        Prepare time series data from DataFrame
        """
        # Validasi panjang prediksi
        if prediction_length > 30:
            st.warning("Prediction length dibatasi maksimal 30 langkah. Akan disesuaikan.")
            prediction_length = min(prediction_length, 30)
        
        # Ensure data is sorted by date
        df = df.sort_values(by=date_column)
        
        # Convert date column to datetime
        df[date_column] = pd.to_datetime(df[date_column])
        
        # Set index to date
        df.set_index(date_column, inplace=True)
        
        # Extract numeric series
        self.original_series = df[value_column].values
        
        # Convert to tensor
        self.context = torch.tensor(self.original_series[-context_length:], dtype=torch.float32)
        
        return self.context, prediction_length

    def forecast(self, context, prediction_length=30, num_samples=100):
        """
        Perform time series forecasting
        """
        # Pastikan prediksi tidak melebihi 30 langkah
        prediction_length = min(prediction_length, 30)
        
        forecasts = self.pipeline.predict(context, prediction_length, num_samples=num_samples)
        return forecasts

    def visualize_forecast(self, forecasts, original_series):
        """
        Create comprehensive visualization of predictions
        """
        plt.figure(figsize=(16, 8), dpi=100, facecolor='white')
        
        # Calculate forecast statistics
        forecast_np = forecasts[0].numpy()
        low, median, high = np.quantile(forecast_np, [0.1, 0.5, 0.9], axis=0)
        
        # Plot original series
        plt.plot(
            range(len(original_series)), 
            original_series, 
            label='Historical Data', 
            color='#2C3E50', 
            linewidth=2,
            alpha=0.7
        )
        
        # Forecast index
        forecast_index = range(len(original_series), len(original_series) + len(median))
        
        # Plot median forecast
        plt.plot(
            forecast_index, 
            median, 
            color='#3498DB', 
            linewidth=3, 
            label='Median Forecast'
        )
        
        # Plot prediction interval
        plt.fill_between(
            forecast_index, 
            low, 
            high, 
            color='#3498DB', 
            alpha=0.2, 
            label='90% Prediction Interval'
        )
        
        plt.title('Advanced Time Series Forecast (Max 30 Steps)', fontsize=18, fontweight='bold', color='#2C3E50')
        plt.xlabel('Time Steps', fontsize=12, color='#34495E')
        plt.ylabel('Value', fontsize=12, color='#34495E')
        plt.legend(frameon=False)
        plt.grid(True, linestyle='--', color='#BDC3C7', alpha=0.5)
        
        # Sophisticated styling
        plt.tight_layout()
        
        return plt

def main():
    # Page configuration
    st.set_page_config(
        page_title="Tempus", 
        page_icon="📊", 
        layout="wide"
    )

    # Modern, minimalist styling
    st.markdown("""
    <style>
    .stApp {
        background-color: #FFFFFF;
        font-family: 'Inter', 'Roboto', sans-serif;
    }
    
    .stButton>button {
        background-color: #3498DB;
        color: white;
        border: none;
        border-radius: 8px;
        padding: 10px 20px;
        transition: all 0.3s ease;
        font-weight: 600;
    }
    
    .stButton>button:hover {
        background-color: #2980B9;
        transform: scale(1.05);
        box-shadow: 0 4px 6px rgba(0,0,0,0.1);
    }
    
    /* Sleek file uploader */
    .stFileUploader {
        background-color: #F8F9FA;
        border: 2px solid #3498DB;
        border-radius: 10px;
        padding: 20px;
        text-align: center;
        transition: all 0.3s ease;
    }
    .stFileUploader:hover {
        border-color: #2980B9;
        background-color: #EAF2F8;
    }
    </style>
    """, unsafe_allow_html=True)

    # Elegant title
    st.markdown(
        "<h1 style='text-align: center; color: #3498DB; margin-bottom: 30px;'>Tempus</h1>", 
        unsafe_allow_html=True
    )

    # File upload
    uploaded_file = st.file_uploader(
        "Upload Time Series Data", 
        type=['csv'], 
        help="CSV with timestamp and numeric columns"
    )

    if uploaded_file is not None:
        # Read CSV
        df = pd.read_csv(uploaded_file)
        
        # Column selection
        col1, col2 = st.columns(2)
        
        with col1:
            date_column = st.selectbox('Date Column', options=df.columns)
        
        with col2:
            value_column = st.selectbox(
                'Value Column', 
                options=[col for col in df.columns if col != date_column]
            )
        
        # Advanced prediction settings
        col3, col4 = st.columns(2)
        
        with col3:
            context_length = st.slider(
                'Historical Context', 
                min_value=30, 
                max_value=500, 
                value=100,
                help="Number of past data points to analyze"
            )
        
        with col4:
            prediction_length = st.slider(
                'Forecast Horizon', 
                min_value=1, 
                max_value=30,  # Dibatasi maksimal 30
                value=30,       # Default 30
                help="Number of future time steps to predict (max 30)"
            )
        
        # Forecast generation
        if st.button('Generate Forecast'):
            try:
                # Initialize and run forecaster
                forecaster = TimeSeriesForecaster()
                
                # Preprocess data
                context, prediction_length = forecaster.preprocess_data(
                    df, 
                    date_column, 
                    value_column, 
                    context_length, 
                    prediction_length
                )
                
                # Perform forecasting
                forecasts = forecaster.forecast(context, prediction_length)
                
                # Visualization
                plt = forecaster.visualize_forecast(forecasts, forecaster.original_series)
                st.pyplot(plt)
                
                # Forecast details
                forecast_np = forecasts[0].numpy()
                forecast_mean = forecast_np.mean(axis=0)
                forecast_lower = np.percentile(forecast_np, 10, axis=0)
                forecast_upper = np.percentile(forecast_np, 90, axis=0)
                
                st.subheader('Forecast Insights')
                prediction_df = pd.DataFrame({
                    'Mean Forecast': forecast_mean,
                    'Lower Bound (10%)': forecast_lower,
                    'Upper Bound (90%)': forecast_upper
                })
                
                st.dataframe(prediction_df)
                
            except Exception as e:
                st.error(f"Forecast generation error: {str(e)}")

if __name__ == '__main__':
    main()