HEARTBEAT Anomaly Detection with LSTM Autoencoder, CNN, Isolation Forest Time Series Python [2024]

Channel	Publish Date	Thumbnail & View Count	Download Video
	Publish Date not found	0 Views	View on YouTube Download Video

#timeseries #lstm #anomalydetection
This is the third video in a series about end-to-end data science projects with machine learning and deep learning. Transform your Data Science skills with this End-to-End Data Science project. I used Kaggle dataset ECG heart rate data for time series analysis and time series anomaly detection. By using the feature extraction of Tensorflow Keras 1. Autoencoders, 2. LSTM Autoencoder and 3. LSTM Autoencoder with Convolutional Neural Network (CNN) in Keras and comparing them with 4. Isolation Forest in ScikitLearn, this project demonstrates advanced machine learning and deep learning techniques for anomaly detection in Python. Perfect for your data science portfolio, this deep learning autoencoder project also explores Area Under Precision and Recall metrics for evaluating model performance.

Get access to my 20 years of experience in AI:

️Free guide: https://www.maryammiradi.com/free-guide
️AI training: https://www.maryammiradi.com/training

Connect with me!

Linkedin ︎ https://www.linkedin.com/in/maryammiradi/

Mentioned in this video

Link to Python code ︎
https://colab.research.google.com/drive/1fp8ZDIXHthPDQp46U0_H5McUWe0tRbfr?uspsharing
Link to Kaggle HeartBeat ECG data ︎
https://www.kaggle.com/datasets/shayanfazeli/heartbeat
Data Science Projects ︎ https://youtube.com/playlist?listPLBYdyi7-DjV33GOnmCbtR9mLOEQr9txnz&siWtisP0V-7dM4Wc_W

Time codes

0:00 Introduction
0:49 Description of Kaggle Data electrocardiogram (ECG) of heartbeats for time series analysis Python
1:16 Python Libraries Keras Tensorflow, StatsModel, ScikitLearn, Pandas, NumPy, Matplotlib, Seaborn
2:15 Exploratory data analysis
3:04 Time series analysis using Statsmodel Python Library using visualization
4:04 Time series analysis using decomposition (trend, time series seasonality and residuals)
5:21 Continuous statistics and how to check whether data is stationary or not
6:42 Improved Dickey-Fuller test
7:17 Autocorrelation function (ACF) and partial autocorrelation function (PACF)
8:28 Anomaly detection Isolation Forest Isolation Forest implementation (incl. sklearn minmaxscaler & forward fill NA) & Results analysis using Confusion Matrix in machine learning
12:06 Deep Learning Autoencoder anomaly detection (Keras Autoencoder), incl. encoder and decoder deep learning with tensorflow keras, building tensorflow keras sequential model and sigmoid function, calculating the reconstruction error
14:48 Long Term Memory LSTM Autoencoder (similar to LLMs) relu activation function with adam optimizer and mean square error loss function. Using the area under precision and recall we compare
the true positive and true negative and classification threshold used
19:40 Get my free guide: 10x Your AI Solutions for the Real World with 10 Steps Data Science Road Map 100 Python Libraries for Impactful Machine Learning and Deep Learning
20:30 LSTM Autoencoder anomaly detection Keras plus Convolutional Neural Network (CNN) Feature Extractor using Convolutional Neural Network and comparing deep learning models using auc pr due imbalanced data
22:00 Max pooling of convolutional neural network
23:54 Comparison of Isolation Forest, AutoEncoder, LSTM Autoencoder and LSTM AutoEncoder Convolutional Neural Network Feature Extraction
24:47 Huperparameter Tuning (Reply if you want a follow-up)
24:58 Off