Deep Learning - LSTM - Financial Time Series Forecasting
Introduction to Long Short Term Memory – LSTM [HANDS ON MODULE]
- RNN model and limitations (vanishing gradient during backpropagation).
- Sequencial data
- LSTM architecture (forget, input and output gates)
- LSTM networks aplication in time series forecasting
- LSTM aplication in NLP
- Predictive modeling
RNN
The capture short term dependencies but struggle capturing long term dependencies.
They capture dependencies maintaining a hidden state that is updated with each input in a sequence (one to timestep to the next).
LSTM
is a RNN that can capture long-term dependencies in sequential data.
They use a memory cell and gates to control the flow of information., allowing to selectively retain of discard information.
TIPES OF GATES (for memory cell)
- Input : with information store/enter in the memory cell.
- Forget : with information is not longer important to discard from memory cell – is trainned-.
- Output : with information to use for the output of the LSTM – is trainned-.
APLICATIONS OF LSTM
- Time series prediction
- Sentimental analyisis
- Voice recognition
- Language simulation
- Hadwirting recognition
- Video analysis
Principles:
- Time series data can’t be randomly split into training and testing; because time series follows a chronological order, and including future data in training leads to data leakage.
The key property that enables Recurrent Neural Networks (RNNs) to handle sequential data is:
Shared weights across time steps
Gates applied sigmon activation to filter information.
PROCESS:
- Packages
- Import data, cleaning and visualization
- Split the data
-
Machine Learning
Feature Engeniering
Deep Learning
CLASS IMBALANCE
https://learn.cqf.com/courses/330/pages/ja25t7-final-project-tutorial-i-2?module_item_id=4399
https://link.springer.com/article/10.1007/s10479-023-05810-8
a
https://journalofbigdata.springeropen.com/articles/10.1186/s40537-020-00333-6
Required packages
-
Resource:
https://github.com/https-deeplearning-ai/tensorflow-1-public
https://www.coursera.org/learn/nlp-sequence-models
