Sequences, Time Series and Prediction Course Syllabus

Overview: This course provides a comprehensive introduction to time series forecasting using TensorFlow, blending theoretical concepts with hands-on practice. Through six structured modules, learners will explore foundational and advanced techniques in sequence modeling, including moving averages, deep neural networks (DNNs), recurrent neural networks (RNNs), and hybrid models. The course emphasizes real-world application, culminating in a project predicting sunspot activity. With approximately 25-30 hours of content, the flexible schedule is designed for working professionals seeking to enhance their machine learning expertise.

Module 1: Sequences and Prediction

Estimated time: 5 hours

• Introduction to time series data and forecasting
• Common patterns in time series: trends, seasonality, and noise
• Evaluation metrics for forecasting models
• Hands-on lab: Time series forecasting with moving averages

Module 2: Deep Neural Networks for Time Series

Estimated time: 5 hours

• Windowing techniques for sequence data
• Preparing features and labels from time series
• Building and training DNNs for forecasting
• Improving model performance with learning rate scheduling

Module 3: Recurrent Neural Networks for Time Series

Estimated time: 5 hours

• Introduction to RNNs for sequential data modeling
• Long Short-Term Memory (LSTM) networks
• Training RNNs and LSTMs on time series data
• Handling long-term dependencies in sequences

Module 4: Real-world Time Series Data

Estimated time: 5 hours

• Challenges in real-world time series datasets
• Preprocessing techniques: differencing and normalization
• Combining CNNs and RNNs for enhanced forecasting

Module 5: Sunspot Activity Prediction Project

Estimated time: 5 hours

• Exploring the sunspot dataset
• Applying windowing, differencing, and feature engineering
• Training and evaluating a combined CNN-RNN model

Module 6: Final Project

Estimated time: 5 hours

• Build a complete time series forecasting pipeline
• Predict sunspot activity using TensorFlow
• Submit model predictions and analysis for evaluation

Prerequisites

• Familiarity with Python programming
• Basic understanding of machine learning concepts
• Experience with TensorFlow or equivalent framework recommended

What You'll Be Able to Do After

• Implement best practices for preparing time series data for machine learning
• Build and train DNNs, RNNs, and LSTMs for forecasting tasks
• Apply moving averages, differencing, and windowing to improve model accuracy
• Develop a real-world prediction model using TensorFlow
• Combine CNNs and RNNs to model complex time series patterns