AWS Certified Machine Learning – SpecialtyModelingEasy
A financial institution is developing a machine learning model to predict stock price movements. They have historical data spanning several years, including daily opening, closing, high, and low prices, as well as trading volume. The data exhibits strong temporal dependencies, where past prices significantly influence future prices. Which type of model architecture is best suited for capturing these sequential patterns for time series forecasting?
- ARecurrent Neural Network (RNN) with LSTM units
- BConvolutional Neural Network (CNN)
- CSupport Vector Machine (SVM)
- DMultilayer Perceptron (MLP)
Show answer & explanationAnswer & explanation
Correct answer: A. Recurrent Neural Network (RNN) with LSTM units
Time series data, like stock prices, inherently has sequential dependencies. RNNs, especially those with LSTM units, are specifically designed to process sequences and maintain memory of past information, making them ideal for such tasks.
Why the other options are wrong
- B. CNNs are primarily designed for spatial hierarchies (like images) and do not inherently handle long-term temporal dependencies well.
- C. SVMs are powerful for classification and regression but are not inherently designed to model temporal dependencies in sequential data effectively.
- D. MLPs treat inputs independently and do not have memory of past inputs, making them unsuitable for sequential data.
Recurrent Neural Networks (RNNs)
A class of neural networks designed to recognize patterns in sequences of data, such as time series, speech, or text.
- Contain loops allowing information to persist.
- LSTMs (Long Short-Term Memory) are a type of RNN that address vanishing gradient problem.
- Suitable for tasks with sequential dependencies.
Memory trick: For sequences, remember the past with RNNs.