Demand Forecast

The Demand Forecast application is a comprehensive AI-powered forecasting platform designed to generate accurate demand predictions from historical sales data, seasonality patterns, and market trends. Built for retail, manufacturing, and supply chain planning, it provides an intuitive interface for creating, managing, and analyzing forecasts.

Overview

The forecast application enables users to:

• Upload and manage datasets containing historical demand data
• Create forecasts using multiple AI models and algorithms
• Run backtests to evaluate model performance on historical data
• Visualize results with interactive charts and detailed analytics
• Export forecasts for integration with other business systems

Key Features

🎯 Multi-Model Forecasting

The platform supports various forecasting algorithms including:

• Exponential Smoothing (ETS) - Traditional time series method
• ARIMA - Autoregressive integrated moving average
• LSTM - Long short-term memory neural networks
• N-BEATS/N-HITS - Neural basis expansion analysis
• PatchTST - Patch-based time series transformer
• Autoformer - Transformer-based forecasting
• LightGBM - Gradient boosting machine
• Mean Ensemble - Combines predictions from all selected models

📊 Backtesting Framework

• Evaluate model performance on historical data
• Compare multiple models simultaneously
• Generate accuracy metrics (WMAPE)
• Identify the best performing models for your data

📈 Interactive Visualizations

• Time series charts showing historical vs forecasted data
• Multi-model comparison views
• Item-level analysis and filtering
• Export capabilities for further analysis

🔄 Workflow Management

• Track process status (Pending, Running, Completed, Failed)
• View recent tasks and their progress
• Manage datasets, forecasts, and backtests

Application Structure

Core Components

1. Dashboard

The main workspace dashboard provides:

• Action Cards: Quick access to create datasets, forecasts, and backtests
• Recent Tasks: Overview of recent processes and their status
• Workspace Information: Current workspace details

2. Datasets

Dataset management functionality:

• Upload Data: Support for CSV, Excel, and JSON files
• Data Types:
  • Main Historical Data: Time series demand data
  • External Data: Price, promotions, inventory levels
  • Categorical Data: Product attributes and metadata
• Data Insights: Automatic analysis of data characteristics
• Item Classification: Identifies new, obsolete, and intermittent items

3. Forecasts

Forecast creation and management:

• Create Forecast: Configure datasets, horizon, frequency, and models
• Forecast Results: Interactive charts and model comparisons
• Export Data: Download forecasts in CSV format
• Model Selection: Choose from available forecasting algorithms

4. Backtests

Model evaluation and comparison:

• Create Backtest: Configure test parameters and model selection
• Performance Metrics: Detailed accuracy analysis
• Model Comparison: Side-by-side evaluation of different algorithms
• Results Visualization: Charts showing historical vs predicted values

Data Flow

Usage Guide

Getting Started

1. Create a Workspace

   • Navigate to the forecast application
   • Create a new workspace or select an existing one

2. Upload Your Data

   • Click "Upload data" on the dashboard
   • Select your dataset type (Main, External, or Categorical)
   • Upload CSV files
   • Configure dataset name and frequency

3. Evaluate with Backtests

   • Create a backtest using the same data
   • Compare model performance
   • Identify the best models for your use case

4. Create Your First Forecast

   • Select your main dataset
   • Choose forecast horizon and frequency
   • Select forecasting models
   • Start the forecast process

Best Practices

Data Preparation

• Ensure consistent date formats (e.g. '2025-07-01')
• Include sufficient historical data (minimum 12-24 periods)
• Clean missing values and outliers
• Use appropriate data frequency (daily, weekly, monthly)

Model Selection

• Test individual models to understand strengths
• Use backtests to validate model choices
• Consider data characteristics when selecting models

Forecast Configuration

• Set realistic forecast horizons
• Monitor forecast accuracy over time

---

The Demand Forecast application provides enterprise-grade forecasting capabilities with an intuitive interface, making advanced AI forecasting accessible to businesses of all sizes.