InsightFlow AI

• Problem Statement

• Supply chain forecasting requires technical expertise

• Raw datasets need extensive preprocessing

• Model selection is often manual and error-prone

• Forecast results are difficult for non-technical users to interpret

• Insights are not easily accessible to decision-makers

  
  

• Proposed Solution

An AI-driven assistant that:

• Understands raw supply chain datasets automatically

• Performs data analysis and feature engineering

• Selects the most suitable forecasting model

• Generates clear visualizations and forecasts

• Explains results using natural language through LLMs

• Provides a seamless, user-friendly interface

  
  

• System Workflow (Project Steps)

1- Dataset Upload

• User uploads a raw supply chain dataset

• No prior data cleaning or technical input required

2- Automated Data Understanding

• Dataset structure and features are analyzed

• Missing values and data types are identified

• Basic statistical summaries are generated

3- Preprocessing & Feature Engineering

• Automated preprocessing pipelines

• Feature extraction and transformation

• Time-series preparation for forecasting

4- Smart Model Selection

• Multiple forecasting models are evaluated

• Automatic selection based on data characteristics and performance

• Supported models include:

  • Prophet

  • Random Forest

  • Gradient Boosting

5- Forecasting & Analysis

• Demand forecasts are generated

• Trend, seasonality, and performance metrics are analyzed

• Results are visualized through interactive dashboards

6- LLM-Powered Explanation

• Large Language Models interpret the results

• Insights are explained in clear, simple language

• Natural-language analytics for non-technical users

7- User Interface & Deployment

• Interactive Streamlit web application

• Containerized deployment using Docker

• Hosted via Render with GitHub version control

  
  

• Technical Highlights

• Large Language Models: Gemini & Llama for natural-language reasoning

• Machine Learning: Prophet, Random Forest, Gradient Boosting

• Automation: Preprocessing & feature engineering pipelines

• Visualization: Interactive dashboards for trends and forecasts

• Deployment: Docker, Render, GitHub