This repository is my suggestion to an anomaly detection exercise within the context of machine learning, specifically for aircraft system data. The dataset provided contains measurements of 11 parameters (e.g., speed, temperature, pressure) across various cycles and time windows. Each data segment is identified by a unique "day_cycle_window" label, referencing the day, cycle, and window of measurements.
Objectives:
- Problem Definition: Formulate a strategy for detecting abnormal windows in the data using the provided dataset.
- Approach Development: Design and implement an algorithm capable of identifying anomalies.
- Results Presentation: Present findings in a way that allows an expert to validate the results.
Key Dataset Details:
- Structure: The dataset consists of multiple parameters recorded across time windows, where each window contains up to 100 data points.
- Goal: Use this information to build an anomaly detection algorithm that flags windows that deviate from normal patterns.