This course, “Time Series Analysis and Forecasting Using Python,” systematically elaborates on the primary notions and methodologies of time series data analysis and forecasting. It starts by defining what time series data is and how to import and manipulate it with the help of SQL Alchemy and Pandas. You will also learn how to change data formats and will be introduced to the DarTS time series class for the initial data analyses with a simple applied example included, such as exponential smoothing.
Following this, the “Time Series Analysis and Forecasting Using Python” course narrows down to the time series data structure with a view on the trends, seasons as well as points of change. While using the Neural Prophet model to make various prophecies concerning different hypothetical cases you wish to study, these components will be explained. Another area to cover within the course is the ARIMA model details that teach you about the ADF test on stationarity, the structures of auto-correlation functions, and the identification of the correct model parameters with the help of the Akaike Information Criterion.
Further, the course provides an understanding of time series forecasting using Supervised Machine Learning, particularly using the Light Gradient Boosting Machine (LightGBM). You will be able to make lagged features for your models. Also, the concept of conformal predictions, and consequently the EnbPI algorithm, and how to compute coverage scores, is discussed in Time Series Analysis and Forecasting Using Python. The last part presents the Lag-Llama model, which is an open-source tool for zero-shot learning in time series forecasting – therefore, completing a set of tools necessary to solve time series issues in Python.
Time Series Analysis and Forecasting Using Python Table of Contents:
- Time Series Analysis and Forecasting Using Python - introductory segment (02:35)
- Time Series data and data generating process (03:08)
- Read, import, and analyze Time Series data - SQLAlchemy, pandas (09:01)
- Long-Form and Wide-Form Time Series Data (04:42)
- DarTS for time series analysis and Preliminary Data Visualizations (07:15)
- Basic Example of Exponential Smoothing using DarTS (05:30)
- Composition of time series - Trend, Seasonality and Change point detection (08:30)
- Set up Google Colab notebook for the analysis of trend and seasonality effects (04:32)
- Investigate scenarios related to Trend, Seasonality Effects and Change points (06:57)
- Investigate scenarios related to Auto-Regressive effects in Neural Prophet (06:45)
- Investigate Effects of Covariates on the forecast predictions in Neural Prophet (05:36)
- Introductory segment on ARIMA (02:19)
- Analysis of Stationarity Effects in Time Series (Statistical test: ADF) (04:48)
- Auto-Correlation Function and Partial Auto-Correlation Function in Time Series (08:46)
- Akaike Information Criterion: ARIMA Model (differencing, MA, and AR parameters) (07:01)
- Introduction to Time Series Forecasting using Supervised Machine Learning (03:11)
- Setting up the Google Colab notebook and Extracting Date Related Features (03:38)
- Creation of Lagged Features for a Time Series Forecasting model (03:33)
- Tree-Based Time Series Forecasting using LightGBM (05:48)
- Conformal Predictions in Time Series Forecasting - Introductory Segment (02:11)
- Exchangeability Hypothesis and Ensemble Batch Prediction Intervals (02:19)
- EnbPI Algorithm Explanation and Setting up Google Colab Notebook (03:45)
- Random Forest Regressor, Mapie Time Series Regressor, and Coverage Score (02:43)
- Introductory Segment on Lag-Llama Model (02:01)
- Applying Language Model such as Lag-Llama for Time Series Forecasting (03:20)
- Zero Shot Generalization capability of the Lag-Llama model & Set up Google Colab (05:31)
- Forecast Predictions and CRPS Evaluation Metric for the Lag-Llama Model (03:29)
Who is this course for?
- This course suits anybody who wants to understand Time Series Analysis and Forecasting.
