AI Project Cycle Overview

AI Project Cycle

• The AI Project Cycle is a step-by-step process to derive value from an AI project and solve problems.
• The cycle has five stages: Problem Scoping, Data Acquisition, Data Preprocessing, Model Training, and Model Evaluation and Deployment.

Problem Scoping

• Problem scoping is defining the project's goal and identifying the problem's different aspects.
• The '4Ws' help identify and understand the problem: Who, What, Where, and Why.
• Who identifies the stakeholders directly and indirectly affected by the problem.
• What analyzes the problem's nature, gathering evidence to confirm its existence.
• Where identifies the situation and where the problem arises.
• Why explains the need to address the problem and the advantages for stakeholders after solving it.
• The problem statement template summarizes the key points gathered from the '4Ws'.

Data Acquisition

• Data acquisition involves collecting accurate and reliable data for analysis.
• Data can be structured, unstructured, or both.
• Structured data is organized in a standardized format, easily accessible by humans and programs. It includes numbers, characters, and special characters.
• Unstructured data doesn't follow traditional data models, making it difficult to store and manage. Examples include videos, audio, images, and log files.
• Dataset is a collection of data in tabular format related to a specific subject.
• Datasets are divided into training and test datasets.
• Training dataset (usually 80%) teaches a machine learning model to make judgments or perform tasks.
• Test dataset (usually 20%) is used to evaluate the model's performance on unseen data.

Acquiring Data from Reliable Sources

• Surveys gather data through questionnaires from a predetermined sample to gain insights.
• Cameras collect visual data (unstructured) that can be analyzed by machine learning.
• Web Scripting collects structured data from the internet, such as market research or price tracking.
• Observation provides data through attentive monitoring and noticing patterns.
• Sensors collect data about physical properties using devices like biometrics.
• Application Program Interface (API) allows two applications to communicate and exchange data.

System Map

• A System Map visually represents the relationships between components within a system.
• The map depicts cause and effect relationships using arrows to connect elements.
• An example is the Water Cycle System Map, where all constituents are circled, and arrows show their interactions.