ANOMALY DETECTION USING LSTM AUTOENCODER

Objective

Goal of this project is to build and train a Long Short Term Memory based Autoencoder model to detect anomalies (i.e. sudden price changes) in the S&P 500 index data.This is implemented using Keras API with Tensorflow 2.0 as backend.

Data

Source - https://www.kaggle.com/pdquant/sp500-daily-19862018

Dataset includes two columns with one containing a daily timestamp and the second with the raw, un-adjusted closing prices for each day.

Requirements

1. numpy
2. pandas
3. Keras
4. Tensorflow 2.0
5. matplotlib
6. seaborn

RESULTS

Autoencoder model was trained for 100 epochs and loss distribution was plotted to obtain a suitable threshold value (0.65) for identifying an anomaly. Test loss ~ 0.098