Complete Technical Paper Content — Ai Port Optimization
1. Executive Abstract, Problem Formulation & Background
AI Framework for Port Optimization This repository contains the implementation of an artificial intelligence framework for enhancing port resilience and intermodal supply chain efficiency, as described in the paper "An Artificial Intelligence Framework for Enhancing U.S. Port Resilience and Intermodal Supply Chain Efficiency." Overview This project demonstrates: XGBoost regression for predicting vessel wait times and truck turn times Digital twin simulation of port operations using discrete-event modeling Integration of machine learning with operational optimization Dataset Due to data use agreements, the original proprietary data from the Port of Los Angeles cannot be shared. However, this repository includes: synthetic_port_data.csv : Generated dataset mimicking real port operations data_generator.py : Script to regenerate synthetic data with similar statistical properties Installation git clone https://github.com/[your-username]/port-ai-optimization.git cd port-ai-optimization pip install -r requirements.txt
This repository contains the implementation of an artificial intelligence framework for enhancing port resilience and intermodal supply chain efficiency, as described in the paper "An Artificial Intelligence Framework for Enhancing U.S. Port Resilience and Intermodal Supply Chain Efficiency."
Due to data use agreements, the original proprietary data from the Port of Los Angeles cannot be shared. However, this repository includes:
git clone https://github.com/[your-username]/port-ai-optimization.git cd port-ai-optimization pip install -r requirements.txt
AI Framework for Port Optimization - Code for the paper "An Artificial Intelligence Framework for Enhancing U.S. Port Resilience and Intermodal Supply Chain Efficiency
Smart Port AI Optimizer Smart Port AI Optimizer is an AI-based system designed to improve port operations through intelligent data analysis, predictive insights, and resource optimization. Project Overview The project focuses on enhancing operational efficiency in ports by analyzing data related to ship movement, berth allocation, truck flow, resource usage, and operational conditions. The system uses artificial intelligence to identify inefficiencies, reduce delays, and support better decision-making. Features Port operations analysis Resource optimization Predictive insights for delays and congestion Smart monitoring dashboard Decision support for port management Technologies Used Python Artificial Intelligence Data Analysis Machine Learning Dashboard Systems Purpose The goal of this project is to support digital transformation in ports and logistics by using AI to improve efficiency, reduce waiting time, and optimize operational performance.
Smart Port AI Optimizer is an AI-based system designed to improve port operations through intelligent data analysis, predictive insights, and resource optimization.
The project focuses on enhancing operational efficiency in ports by analyzing data related to ship movement, berth allocation, truck flow, resource usage, and operational conditions. The system uses artificial intelligence to identify inefficiencies, reduce delays, and support better decision-making.
The goal of this project is to support digital transformation in ports and logistics by using AI to improve efficiency, reduce waiting time, and optimize operational performance.
AI-based smart port optimization system for improving port operations, reducing delays, and enhancing resource management through data analysis and predictive insights.
2. Related Work, Literature Survey & Algorithmic Foundations
Features Reinforcement learning-based optimization Digital twin simulation Workflow analysis Resource allocation optimization Automated testing environment Tech Stack Python JavaScript HTML/CSS AI/ML Concepts Future Improvements Integrate live IoT sensor streams Cloud deployment using Docker and Kubernetes Predictive maintenance module Real-time analytics dashboard Multi-port support
AI-powered smart port optimization system using simulation, workflow analysis, and intelligent automation.
🧠 IBM Deep Learning Engineering PyTorch • Keras • TensorFlow • CNNs • Transformers • Computer Vision • Deep Learning A hands-on deep learning engineering portfolio built through the IBM Deep Learning with PyTorch, Keras and TensorFlow Professional Certificate. This repository contains my implementations, experiments, notebooks, projects, and engineering notes covering the complete deep learning workflow — from neural-network fundamentals and optimization to advanced architectures, computer vision, transformers, generative models, and an end-to-end AI capstone. The goal is not only to understand deep learning theory, but to demonstrate the ability to build, train, evaluate, optimize, and apply neural networks to practical problems. 🚀 What This Repository Demonstrates Area Technologies / Concepts 🧠 Deep Learning Neural Networks, Backpropagation, Optimization 🔥 PyTorch Tensors, Autograd, Custom Models, Training Pipelines 🤖 TensorFlow TensorFlow 2, Model Training, Deployment Concepts 🧩 Keras Custom Layers, Models, Callbacks, Training 👁️ Computer Vision CNNs, Image Classification, Transfer Learning 🔄 Sequence Modeling RNNs, Sequential Data, Time-Series 🤯 Transformers Attention, Transformer Architectures 🎨 Generative AI Autoencoders, GANs 🎯 Reinforcement Learning RL Fundamentals and Neural Approaches 📊 Machine Learning Regression, Classification, Gradient Descent ⚡ Model Optimization Training Optimization, Evaluation, Experimentation 🧪 Model Evaluation Accuracy, Loss, Generalization, Performance Analysis 🛠️ Engineering Python, NumPy, Matplotlib, Jupyter 📚 Learning Path The repository follows a structured progression through five major stages. DEEP LEARNING ENGINEERING │ ▼ ┌─────────────────────────────┐ │ 1. Neural Network Basics │ │ Regression / Classification│ │ Gradient Descent │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 2. Keras + TF │ │ Custom Models / Layers │ │ CNNs / Training Pipelines │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 3. PyTorch │ │ Tensors / Autograd │ │ Neural Network Engineering │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 4. Advanced DL │ │ CNNs / Transformers │ │ Transfer Learning / GANs │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 5. AI Capstone │ │ End-to-End DL Application │ │ Evaluation + Analysis │ └─────────────────────────────┘ 📂 Repository Structure ibm-deep-learning-engineering/ ├── 01-introduction-deep-learning-keras/ │ ├── exercises/ │ ├── experiments/ │ ├── notebooks/ │ └── README.md ├── 02-deep-learning-keras-tensorflow/ │ ├── cnn/ │ ├── custom-models/ │ ├── experiments/ │ ├── notebooks/ │ └── README.md ├── 03-neural-networks-pytorch/ │ ├── autograd/ │ ├── neural-networks/ │ ├── notebooks/ │ ├── tensors/ │ ├── training/ │ └── README.md ├── 04-advanced-deep-learning-pytorch/ │ ├── autoencoders/ │ ├── cnn/ │ ├── gan/ │ ├── sequence-models/ │ ├── transfer-learning/ │ ├── transformers/ │ └── README.md ├── 05-ai-capstone/ │ ├── data/ │ ├── evaluation/ │ ├── models/ │ ├── notebooks/ │ ├── results/ │ ├── src/ │ └── README.md ├── assets/ │ ├── architecture/ │ ├── confusion-matrix/ │ ├── results/ │ └── training-curves/ ├── environment.yml ├── README.md └── requirements.txt Folders currently hold placeholder .gitkeep files and will be populated as each course's notebooks and experiments are added. 🔥 Core Topics 1️⃣ Neural Network Fundamentals Implementation and experimentation with the fundamental building blocks of deep learning. Concepts Artificial Neural Networks Neurons and Layers Activation Functions Forward Propagation Backpropagation Gradient Descent Loss Functions Optimization Regression Logistic Regression Classification Model Training Validation Example Input │ ▼ ┌──────────────┐ │ Input Layer │ └──────┬───────┘ │ ▼ ┌──────────────┐ │ Hidden Layer │ │ ReLU │ └──────┬───────┘ │ ▼ ┌──────────────┐ │ Hidden Layer │ │ ReLU │ └──────┬───────┘ │ ▼ ┌──────────────┐ │ Output Layer │ │ Softmax │ └──────┬───────┘ │ ▼ Prediction 🔥 2️⃣ Keras & TensorFlow Hands-on development using the Keras API and TensorFlow ecosystem. Covered Areas Sequential Models Functional API Custom Layers Custom Models Activation Functions Loss Functions Optimizers Callbacks Model Evaluation CNN architectures Image classification TensorFlow 2 Model experimentation Example workflow: Dataset ↓ Preprocessing ↓ Train / Validation Split ↓ Keras Model ↓ Training ↓ Validation ↓ Evaluation ↓ Prediction 🔥 3️⃣ PyTorch Deep learning implementations using PyTorch. Topics PyTorch Tensors Tensor Operations Autograd Computational Graphs torch.nn Custom Neural Networks Optimizers Loss Functions Training Loops Validation Model Evaluation Dataset / DataLoader GPU acceleration concepts Typical PyTorch pipeline: Dataset ↓ DataLoader ↓ Model ↓ Forward Pass ↓ Loss ↓ Backward Pass ↓ Optimizer ↓ Parameter Update ↓ Evaluation 👁️ 4️⃣ Computer Vision & CNNs Implementation and experimentation with convolutional neural networks. Topics Convolution Kernels / Filters Feature Maps Pooling Stride Padding Receptive Fields CNN Architecture Image Classification Transfer Learning Data Augmentation Model Evaluation Example CNN architecture: Image │ ▼ Conv2D │ ▼ ReLU │ ▼ Pooling │ ▼ Conv2D │ ▼ ReLU │ ▼ Pooling │ ▼ Flatten │ ▼ Dense │ ▼ Softmax │ ▼ Prediction 🤖 5️⃣ Advanced Deep Learning The advanced section explores architectures beyond traditional feed-forward networks. Topics Advanced CNNs Transfer Learning Autoencoders Generative Adversarial Networks Sequence Models Time-Series Prediction Transformer Models Attention Mechanisms Unsupervised Learning Reinforcement Learning 🧩 Transformers A dedicated section explores transformer-based architectures and their role in modern AI. Input Sequence │ ▼ Tokenization │ ▼ Embeddings │ ▼ Positional Information │ ▼ Self-Attention │ ▼ Feed Forward Network │ ▼ Transformer Blocks │ ▼ Output Key concepts: Attention Self-Attention Query / Key / Value Positional Encoding Transformer Blocks Sequence Modeling 🎨 Generative Deep Learning Exploration of generative neural-network architectures. Autoencoders Input │ ▼ Encoder │ ▼ Latent Representation │ ▼ Decoder │ ▼ Reconstructed Output GANs ┌──────────────┐ │ Generator │ └──────┬───────┘ │ ▼ Fake Samples │ ▼ ┌──────────────┐ │ Discriminator│ └──────┬───────┘ │ ▼ Real / Fake 🧪 Model Evaluation A major focus of this repository is understanding why a model performs the way it does , rather than looking only at accuracy. Evaluation includes: Training Loss Validation Loss Accuracy Precision Recall F1 Score Confusion Matrix Overfitting Analysis Underfitting Analysis Generalization Model Comparison Error Analysis Example: MODEL EVALUATION │ ┌────────────┼────────────┐ ▼ ▼ ▼ Accuracy Loss Confusion │ │ Matrix ▼ ▼ ▼ Precision Recall Error Analysis │ │ │ └────────────┼────────────┘ ▼ Deployment Decision 🚀 AI Capstone Project The final project brings together the concepts learned throughout the program into an end-to-end deep learning workflow. Project Pipeline Problem Definition ↓ Data Collection ↓ Data Exploration ↓ Data Preprocessing ↓ Model Design ↓ Training ↓ Hyperparameter Tuning ↓ Evaluation ↓ Error Analysis ↓ Visualization ↓ Final Model ↓ Results & Conclusions The capstone emphasizes not just model development, but also: Experimental methodology Performance analysis Visualization Model comparison Error analysis Technical communication 🛠️ Technology Stack Programming Deep Learning Data & Visualization Development 📊 Skills Demonstrated This repository demonstrates practical experience with: Deep Learning ├── Neural Networks ├── Backpropagation ├── Gradient Descent ├── Optimization ├── CNNs ├── Transfer Learning ├── Transformers ├── Autoencoders ├── GANs ├── Reinforcement Learning └── Sequence Models Frameworks ├── PyTorch ├── TensorFlow └── Keras Engineering ├── Python ├── Model Training ├── Experimentation ├── Evaluation ├── Error Analysis └── Visualization 🎯 Engineering Perspective This repository is designed around an engineering mindset: Understand → Implement → Experiment → Evaluate → Optimize → Apply Instead of treating deep learning as a collection of APIs, each project focuses on understanding the relationship between: Data → Architecture → Training → Optimization → Evaluation → Performance This approach is particularly important when moving from academic deep learning toward practical AI/ML engineering and Edge AI deployment . 📈 Future Extensions The concepts learned here can be extended toward production-oriented AI systems. Planned areas include: ONNX model export ONNX Runtime inference Model quantization INT8 inference Neural-network optimization Hardware acceleration GPU inference Qualcomm AI acceleration Edge AI deployment Embedded Linux AI inference Android AI inference AI performance profiling Model latency optimization Deep Learning ↓ Model Optimization ↓ ONNX ↓ Quantization ↓ Hardware Acceleration ↓ Edge AI ↓ Embedded AI Systems 🏆 Certificate IBM Deep Learning with PyTorch, Keras and TensorFlow Professional Certificate 5-course professional certificate covering: Introduction to Deep Learning & Neural Networks with Keras Deep Learning with Keras and TensorFlow Introduction to Neural Networks and PyTorch Deep Learning with PyTorch AI Capstone Project with Deep Learning 📌 Why This Repository Exists The purpose of this repository is to create a public, continuously evolving portfolio demonstrating practical deep learning engineering skills. Rather than presenting the certificate alone, this repository documents: What I learned How I implemented it How the models work How different architectures behave How models are evaluated What experiments were performed What limitations were discovered How the knowledge can be extended toward production and Edge AI 👨💻 About I am an Embedded Linux / Android Automotive / AI Engineering professional interested in the intersection of: Embedded Systems + Linux / Android + AI / ML + Deep Learning + Hardware Acceleration + Edge AI My long-term focus is building AI-powered embedded and edge computing systems , combining software engineering, deep learning, model optimization, and hardware acceleration. ⭐ Repository Goals Build strong foundations in deep learning Develop practical PyTorch expertise Develop practical TensorFlow/Keras expertise Understand modern neural architectures Build demonstrable AI projects Develop model optimization skills Transition deep learning models toward Edge AI Build production-oriented AI engineering expertise 📜 License This repository is intended for educational and portfolio purposes. Course material and proprietary content remain the property of their respective owners. Original implementations, experiments, and documentation in this repository are provided for learning and demonstration purposes. ⭐ If you find this repository useful Consider giving it a ⭐ and following the repository as it evolves from deep learning fundamentals toward production-grade Edge AI engineering .
A hands-on deep learning engineering portfolio built through the IBM Deep Learning with PyTorch, Keras and TensorFlow Professional Certificate.
This repository contains my implementations, experiments, notebooks, projects, and engineering notes covering the complete deep learning workflow — from neural-network fundamentals and optimization to advanced architectures, computer vision, transformers, generative models, and an end-to-end AI capstone.
The goal is not only to understand deep learning theory, but to demonstrate the ability to build, train, evaluate, optimize, and apply neural networks to practical problems.
DEEP LEARNING ENGINEERING │ ▼ ┌─────────────────────────────┐ │ 1. Neural Network Basics │ │ Regression / Classification│ │ Gradient Descent │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 2. Keras + TF │ │ Custom Models / Layers │ │ CNNs / Training Pipelines │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 3. PyTorch │ │ Tensors / Autograd │ │ Neural Network Engineering │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 4. Advanced DL │ │ CNNs / Transformers │ │ Transfer Learning / GANs │ └──────────────┬──────────────┘ │ ▼ ┌─────────────────────────────┐ │ 5. AI Capstone │ │ End-to-End DL Application │ │ Evaluation + Analysis │ └─────────────────────────────┘
ibm-deep-learning-engineering/ ├── 01-introduction-deep-learning-keras/ │ ├── exercises/ │ ├── experiments/ │ ├── notebooks/ │ └── README.md ├── 02-deep-learning-keras-tensorflow/ │ ├── cnn/ │ ├── custom-models/ │ ├── experiments/ │ ├── notebooks/ │ └── README.md ├── 03-neural-networks-pytorch/ │ ├── autograd/ │ ├── neural-networks/ │ ├── notebooks/ │ ├── tensors/ │ ├── training/ │ └── README.md ├── 04-advanced-deep-learning-pytorch/ │ ├── autoencoders/ │ ├── cnn/ │ ├── gan/ │ ├── sequence-models/ │ ├── transfer-learning/ │ ├── transformers/ │ └── README.md ├── 05-ai-capstone/ │ ├── data/ │ ├── evaluation/ │ ├── models/ │ ├── notebooks/ │ ├── results/ │ ├── src/ │ └── README.md ├── assets/ │ ├── architecture/ │ ├── confusion-matrix/ │ ├── results/ │ └── training-curves/ ├── environment.yml ├── README.md └── requirements.txt
Folders currently hold placeholder .gitkeep files and will be populated as each course's notebooks and experiments are added.
Implementation and experimentation with the fundamental building blocks of deep learning.
3. Dataset Ingestion, Preprocessing & Stochastic Augmentations
Input │ ▼ ┌──────────────┐ │ Input Layer │ └──────┬───────┘ │ ▼ ┌──────────────┐ │ Hidden Layer │ │ ReLU │ └──────┬───────┘ │ ▼ ┌──────────────┐ │ Hidden Layer │ │ ReLU │ └──────┬───────┘ │ ▼ ┌──────────────┐ │ Output Layer │ │ Softmax │ └──────┬───────┘ │ ▼ Prediction
Dataset ↓ Preprocessing ↓ Train / Validation Split ↓ Keras Model ↓ Training ↓ Validation ↓ Evaluation ↓ Prediction
Dataset ↓ DataLoader ↓ Model ↓ Forward Pass ↓ Loss ↓ Backward Pass ↓ Optimizer ↓ Parameter Update ↓ Evaluation
Image │ ▼ Conv2D │ ▼ ReLU │ ▼ Pooling │ ▼ Conv2D │ ▼ ReLU │ ▼ Pooling │ ▼ Flatten │ ▼ Dense │ ▼ Softmax │ ▼ Prediction
A dedicated section explores transformer-based architectures and their role in modern AI.
Input Sequence │ ▼ Tokenization │ ▼ Embeddings │ ▼ Positional Information │ ▼ Self-Attention │ ▼ Feed Forward Network │ ▼ Transformer Blocks │ ▼ Output
Input │ ▼ Encoder │ ▼ Latent Representation │ ▼ Decoder │ ▼ Reconstructed Output
┌──────────────┐ │ Generator │ └──────┬───────┘ │ ▼ Fake Samples │ ▼ ┌──────────────┐ │ Discriminator│ └──────┬───────┘ │ ▼ Real / Fake
A major focus of this repository is understanding why a model performs the way it does , rather than looking only at accuracy.
MODEL EVALUATION │ ┌────────────┼────────────┐ ▼ ▼ ▼ Accuracy Loss Confusion │ │ Matrix ▼ ▼ ▼ Precision Recall Error Analysis │ │ │ └────────────┼────────────┘ ▼ Deployment Decision
4. Neural Network Backbone, Layer Specifications & Loss Formulations
The final project brings together the concepts learned throughout the program into an end-to-end deep learning workflow.
Problem Definition ↓ Data Collection ↓ Data Exploration ↓ Data Preprocessing ↓ Model Design ↓ Training ↓ Hyperparameter Tuning ↓ Evaluation ↓ Error Analysis ↓ Visualization ↓ Final Model ↓ Results & Conclusions
Deep Learning ├── Neural Networks ├── Backpropagation ├── Gradient Descent ├── Optimization ├── CNNs ├── Transfer Learning ├── Transformers ├── Autoencoders ├── GANs ├── Reinforcement Learning └── Sequence Models Frameworks ├── PyTorch ├── TensorFlow └── Keras Engineering ├── Python ├── Model Training ├── Experimentation ├── Evaluation ├── Error Analysis └── Visualization
Instead of treating deep learning as a collection of APIs, each project focuses on understanding the relationship between:
This approach is particularly important when moving from academic deep learning toward practical AI/ML engineering and Edge AI deployment .
Deep Learning ↓ Model Optimization ↓ ONNX ↓ Quantization ↓ Hardware Acceleration ↓ Edge AI ↓ Embedded AI Systems
The purpose of this repository is to create a public, continuously evolving portfolio demonstrating practical deep learning engineering skills.
I am an Embedded Linux / Android Automotive / AI Engineering professional interested in the intersection of:
Embedded Systems + Linux / Android + AI / ML + Deep Learning + Hardware Acceleration + Edge AI
My long-term focus is building AI-powered embedded and edge computing systems , combining software engineering, deep learning, model optimization, and hardware acceleration.
5. Training Protocol, Convergence Behavior & Experimental Evaluation
Course material and proprietary content remain the property of their respective owners. Original implementations, experiments, and documentation in this repository are provided for learning and demonstration purposes.
Consider giving it a ⭐ and following the repository as it evolves from deep learning fundamentals toward production-grade Edge AI engineering .
Hands-on deep learning engineering portfolio covering Keras, TensorFlow, PyTorch, CNNs, Transformers, transfer learning, autoencoders, GANs, reinforcement learning, and an end-to-end AI capstone.
🚢 Smart Port AI Intelligence Platform 📌 Overview The Smart Port AI Intelligence Platform is an industry-grade, end-to-end machine learning ecosystem designed to predict vessel arrival delays, optimize berth assignments, and monitor fleet logistics in real-time. It integrates Machine Learning, FastAPI web services, and a premium glassmorphic Streamlit command center with a production-ready containerized architecture suitable for AI engineering and MLOps workflows. 🎯 Key Features ✔ AI Delay Prediction: Runs a Random Forest Regressor model predicting arrival delays based on vessel dimensions, cargo weights, weather conditions, traffic congestion, and voyage distance. ✔ Smart Berth Allocation Engine: A priority scheduler that load-balances incoming vessel ETAs and dimensions to optimize berth assignments ( B01 to B12 ) in real-time. ✔ Vessel Tracker & GIS Map: An interactive geospatial world map plotting real-time transits of incoming vessels color-coded by AI-predicted risk levels. ✔ Weather Stress Simulator: A stress-testing sandbox that permits operators to override global weather severity (storms) or port traffic levels and view immediate predicted delay impacts. ✔ MLOps Metrics Deck: An analytics section displaying live model parameters, feature importances, and regression performance metrics ( $R^2$ , MAE, RMSE). ✔ Containerized Microservices: Docker configuration with Dockerfile s and docker-compose.yml to run the frontend and backend together seamlessly. ⚙️ Tech Stack Core Language: Python 🐍 ML & Processing: Scikit-learn, Pandas, NumPy Backend API: FastAPI, Uvicorn, Pydantic, pytest, httpx Frontend UI: Streamlit, Plotly (Timelines, Gauges, GIS Scatter maps), Custom CSS (Glassmorphism theme) DevOps: Docker, Docker Compose 📁 Project Structure Smart-Port-AI-Intelligence-Platform/ │ ├── api/ # FastAPI backend │ ├── app.py # API main entrypoint │ ├── routes.py # API routing definitions │ ├── schemas.py # Pydantic validation schemas │ └── services.py # Allocation algorithms & ML interface │ ├── src/ # Machine Learning Pipeline │ ├── models/ # Model weights (.pkl) & metadata (.json) │ ├── inference/ # Predictor class for query inputs │ └── training/ # Model training scripts │ ├── dashboard/ # Streamlit multi-page UI │ ├── app.py # Dashboard homepage │ ├── pages/ # Overview, Predict, Analytics, Tracker, Simulator │ └── theme/ # Custom glassmorphic CSS styling │ ├── data/ # Data storage │ └── raw/generate_data.py # Synthetic data generator │ ├── tests/ # Integration & API tests │ └── test_api.py # Pytest routes checker │ ├── Dockerfile.api # API backend container configuration ├── Dockerfile.dashboard # Streamlit frontend container configuration ├── docker-compose.yml # Multi-container conductor file ├── requirements.txt # Project dependencies └── README.md # Project documentation ▶️ How to Run the Project Option A: Using Docker Compose (Recommended) You can build and spin up the entire containerized platform (API + Frontend) in one command: docker-compose up --build FastAPI Backend: Runs at http://localhost:8000 (API Docs at http://localhost:8000/docs ) Streamlit Dashboard: Runs at http://localhost:8501 Option B: Local Terminal Setup 1. Setup Virtual Environment & Dependencies # Create and activate environment python -m venv venv venv \S cripts \a ctivate # Windows PowerShell # source venv/bin/activate # macOS / Linux # Install packages pip install -r requirements.txt 2. Generate Data & Train Model python data/raw/generate_data.py python src/training/train.py 3. Run FastAPI Backend uvicorn api.app:app --reload 4. Run Streamlit Dashboard (In a new terminal with venv activated) streamlit run dashboard/app.py 🧪 Running Automated Tests Run integration tests to verify endpoint routing and predictive stability: venv/Scripts/python -m pytest 👨💻 Author Shabab Hanaan Software Engineering Student | AI & Machine Learning Enthusiast
The Smart Port AI Intelligence Platform is an industry-grade, end-to-end machine learning ecosystem designed to predict vessel arrival delays, optimize berth assignments, and monitor fleet logistics in real-time.
It integrates Machine Learning, FastAPI web services, and a premium glassmorphic Streamlit command center with a production-ready containerized architecture suitable for AI engineering and MLOps workflows.
✔ AI Delay Prediction: Runs a Random Forest Regressor model predicting arrival delays based on vessel dimensions, cargo weights, weather conditions, traffic congestion, and voyage distance. ✔ Smart Berth Allocation Engine: A priority scheduler that load-balances incoming vessel ETAs and dimensions to optimize berth assignments ( B01 to B12 ) in real-time. ✔ Vessel Tracker & GIS Map: An interactive geospatial world map plotting real-time transits of incoming vessels color-coded by AI-predicted risk levels. ✔ Weather Stress Simulator: A stress-testing sandbox that permits operators to override global weather severity (storms) or port traffic levels and view immediate predicted delay impacts. ✔ MLOps Metrics Deck: An analytics section displaying live model parameters, feature importances, and regression performance metrics ( $R^2$ , MAE, RMSE). ✔ Containerized Microservices: Docker configuration with Dockerfile s and docker-compose.yml to run the frontend and backend together seamlessly.
Smart-Port-AI-Intelligence-Platform/ │ ├── api/ # FastAPI backend │ ├── app.py # API main entrypoint │ ├── routes.py # API routing definitions │ ├── schemas.py # Pydantic validation schemas │ └── services.py # Allocation algorithms & ML interface │ ├── src/ # Machine Learning Pipeline │ ├── models/ # Model weights (.pkl) & metadata (.json) │ ├── inference/ # Predictor class for query inputs │ └── training/ # Model training scripts │ ├── dashboard/ # Streamlit multi-page UI │ ├── app.py # Dashboard homepage │ ├── pages/ # Overview, Predict, Analytics, Tracker, Simulator │ └── theme/ # Custom glassmorphic CSS styling │ ├── data/ # Data storage │ └── raw/generate_data.py # Synthetic data generator │ ├── tests/ # Integration & API tests │ └── test_api.py # Pytest routes checker │ ├── Dockerfile.api # API backend container configuration ├── Dockerfile.dashboard # Streamlit frontend container configuration ├── docker-compose.yml # Multi-container conductor file ├── requirements.txt # Project dependencies └── README.md # Project documentation
You can build and spin up the entire containerized platform (API + Frontend) in one command:
# Create and activate environment python -m venv venv venv \S cripts \a ctivate # Windows PowerShell # source venv/bin/activate # macOS / Linux # Install packages pip install -r requirements.txt
Model Figures & Evaluation Curves (4 Images)
Deep Learning Execution Pipeline
(Preprocessing)
(Feature Extraction)
(Self-Attention)
(Loss Optimization)