EDAN70 Project in Computer Science - DOmingo Carvajal

Anomaly Detection in time series using linear autoencoder.

The source code is in the jupyter notebook.

The anomalies found are in the folder images.

In the folder normal_images are the sequences that were not found anomalies (Only a fraction of them).

In the archives folder there are the results of different variations of the model. m1 indicates the original version of the model with layers of size 14-7-4. m2 is the final model using all the data, and m3 is the final model using the reduced version of the data without the extremes. What is in the folder images and normal images are the results of m3.