Capstone Project in AI Concepts

Capstone Project

• A capstone project is a final project in an academic program. It combines learning from all the program.
• Students look at real-world examples and discuss possible solutions.
• Students learn to apply their knowledge to real-world problems, express solutions non-technically, and select appropriate algorithms.
• Capstone projects are research-based, completed independently, and intended to provide deep understanding of the subject. Students integrate knowledge to demonstrate comprehensive understanding of subjects through a project.

AI Project Cycle Concepts

• Key concepts for AI projects: Project Cycle, Model Validation, RMSE, MSE, MAPE.
• Students must thoroughly research the topic and demonstrate a deep comprehension.

Capstone Project Ideas

• Stock Prices Predictor
• Sentiment Analyzer
• Movie Ticket Price Predictor
• Student Result Predictor
• Human Activity Recognition using Smartphone Data
• Classifying humans and animals in a photo

Understanding the Problem

• Artificial intelligence is one of the most transformative technologies.
• Six steps in every AI project are identifying the problem, gathering data, defining features, constructing AI models, evaluating and refining, and deployment.
• The first step in tackling an AI project is understanding the problem by asking if there is a pattern in the data. Machine learning requires a pattern. If no pattern exists, then the problem cannot be solved using AI technology.

Design Thinking Framework

• Design Thinking is a problem-solving method emphasizing solutions and tackling ill-defined or unknown problems.
• The five stages are: Empathize, Define, Ideate, Prototype, Test. (EDIPT)
• Capstone projects often require using design thinking to approach challenging problems.

Problem Decomposition

• Real complex tasks require breaking down the problem into smaller units.
• Understand the problem, restate it in own words, find desired inputs and outputs, break down into large pieces and then into smaller pieces.
• Code each piece individually and test each individual piece.

Time Series Data Decomposition

• Time series can be broken down into Level, Trend, Seasonality, and Noise components.
• This decomposition helps in understanding and forecasting time series data.
• The Airline Passengers dataset, for example, shows the total number of airline passengers monthly over a period of time.

Analytic Approach

• There are different approaches to solving problems depending on the question being asked.
• Descriptive, Diagnostic, Predictive, and Prescriptive approaches.

Data Requirements

• Defining the necessary data content, format, and sources for initial data collection is critical.
• Determine if more or less data is required to sufficiently address the underlying problem.

Modeling and Evaluation

• Data modeling focuses on creating descriptive or predictive models.
• Creating accurate models requires careful consideration of potential errors and appropriate adjustments.
• Evaluating models involves using a train/test split approach to improve model performance. Train on 80% of the data and test on 20% of the data. This technique allows testing how effective a model is on unseen data. Use mean absolute error (MAE) or root mean squared errors (RMSE) to evaluate the models.

Explore the essentials of undertaking a capstone project focused on AI. This quiz covers various key concepts in the AI project cycle, including model validation and error metrics. Students will also examine innovative project ideas and their real-world applications in AI.