Mathnal Academy — Supply Chain Analytics Courses
Back to Library
Forecasting with Python Verson-1
Forecasting
Bestseller

Forecasting with Python Verson-1

by Krish Naidu

81 pages English Version 1
₹1,000+ 18% GST
Secure online reader — no downloads. Access anytime from your dashboard.

About This Book

TABLE OF CONTENTS Chapter 1 Python for Time Series — The Complete Toolkit 1.1 Time Series Data: What It Is and Why It's Special 1.2 Loading and Indexing Time Series with Pandas 1.3 Resampling, Rolling Windows, and Shifting 1.4 Handling Missing Values and Outliers 1.5 Visualizing Time Series Professionally Chapter 2 Exploratory Data Analysis for Forecasting 2.1 Decomposition: Trend, Seasonality, and Residuals 2.2 Autocorrelation and Partial Autocorrelation (ACF/PACF) 2.3 Stationarity: The Most Important Concept in Time Series 2.4 Unit Root Tests: ADF and KPSS 2.5 Seasonality Detection: FFT and Periodogram Chapter 3 Time Series Data Preparation 3.1 Train/Test Splits for Time Series 3.2 Cross-Validation for Time Series: TimeSeriesSplit 3.3 Feature Engineering from Timestamps 3.4 Normalization and Differencing 3.5 Handling Multiple Seasonalities Chapter 4 Classical Methods: Smoothing and Baseline Models 4.1 Naive Forecasting Methods 4.2 Moving Averages: SMA, WMA, EMA 4.3 Simple Exponential Smoothing (SES) 4.4 Holt's Double Exponential Smoothing (Trend) 4.5 Holt-Winters Triple Exponential Smoothing (Trend + Seasonality) 4.6 Implementing with Statsmodels Chapter 5 ARIMA and SARIMA Models 5.1 The AR, MA, and ARIMA Models — Full Derivation 5.2 Identifying p, d, q from ACF and PACF 5.3 Fitting and Diagnosing ARIMA Models 5.4 SARIMA: Adding Seasonal Components 5.5 Auto-ARIMA with pmdarima 5.6 ARIMAX: Including Exogenous Variables Chapter 6 Machine Learning for Time Series 6.1 Reformulating Forecasting as Supervised Learning 6.2 Feature Engineering: Lags, Rolling Stats, Calendar Features 6.3 XGBoost for Time Series 6.4 LightGBM for Large-Scale Forecasting 6.5 Random Forest for Forecasting 6.6 Feature Importance and Model Interpretation Chapter 7 Prophet: Scalable Forecasting for Business 7.1 Prophet's Decomposition Model 7.2 Fitting Prophet in Python 7.3 Adding Holidays and Special Events 7.4 Custom Seasonalities 7.5 Hyperparameter Tuning with Cross-Validation Chapter 8 Deep Learning for Time Series 8.1 Why Deep Learning for Forecasting? 8.2 LSTM: Long Short-Term Memory Networks 8.3 Building an LSTM in TensorFlow/Keras 8.4 Temporal Fusion Transformer (TFT) Concepts 8.5 Practical Tips: Seq2Seq and Multi-Step Forecasting 8.6 When NOT to Use Deep Learning Chapter 9 Forecast Uncertainty: Prediction Intervals 9.1 Point Forecasts vs. Prediction Intervals 9.2 Conformal Prediction: Distribution-Free Intervals 9.3 Bootstrapped Prediction Intervals 9.4 Quantile Regression for Probabilistic Forecasting 9.5 Fan Charts and Scenario Forecasting Chapter 10 Bayesian Forecasting with PyMC 10.1 Why Bayesian? Prior Knowledge + Data 10.2 The Bayesian Workflow 10.3 Bayesian Linear Trend Model 10.4 Bayesian Structural Time Series 10.5 MCMC Sampling with PyMC 10.6 Posterior Predictive Checks Chapter 11 Forecast Evaluation and Model Selection 11.1 Error Metrics: MAE, MAPE, RMSE, MASE, sMAPE 11.2 Residual Diagnostics 11.3 Time Series Cross-Validation (Walk-Forward) 11.4 Diebold-Mariano Test: Comparing Two Models 11.5 Ensemble Forecasting 11.6 Forecast Combination Strategies Chapter 12 Hierarchical and Multi-Step Forecasting 12.1 Hierarchical Time Series Structures 12.2 Bottom-Up, Top-Down, Middle-Out Reconciliation 12.3 Optimal Reconciliation (MinTrace) 12.4 Direct, Recursive, and MIMO Multi-Step Forecasting 12.5 Global vs. Local Models for Multiple Series Chapter 13 Complete Industry Projects 13.1 Project A: Retail Demand Forecasting (FMCG) 13.2 Project B: Energy Load Forecasting 13.3 Project C: Financial Volatility Forecasting 13.4 Project D: Multi-SKU Hierarchical Forecasting Pipeline

What You'll Learn

Learning Outcomes
By the end of this book, you will be able to:
Understand the fundamentals of time series data and distinguish it from traditional cross-sectional data for forecasting applications.
Prepare and preprocess time series datasets using Python, including indexing, resampling, handling missing values, outlier treatment, and feature engineering.
Perform comprehensive exploratory time series analysis by identifying trends, seasonality, cyclicality, and stationarity using statistical and visual techniques.
Apply statistical tests such as ADF and KPSS to evaluate stationarity and determine appropriate data transformations.
Develop baseline forecasting models using naïve methods, moving averages, and exponential smoothing techniques for benchmarking model performance.
Build and optimize ARIMA, SARIMA, and ARIMAX models, including parameter selection, diagnostics, and incorporation of external variables.
Transform forecasting problems into supervised machine learning tasks using lag variables, rolling statistics, and calendar-based feature engineering.
Develop machine learning forecasting models using Random Forest, XGBoost, and LightGBM, and interpret model behavior through feature importance analysis.
Implement Meta Prophet for business forecasting, including holiday effects, custom seasonalities, and cross-validation for hyperparameter optimization.
Design deep learning forecasting solutions using LSTM networks and understand advanced architectures such as Temporal Fusion Transformers (TFT) for sequential prediction problems.
Construct probabilistic forecasting models by generating prediction intervals using conformal prediction, bootstrapping, and quantile regression techniques.
Develop Bayesian forecasting models using PyMC, incorporating prior knowledge, posterior inference, Bayesian Structural Time Series, and uncertainty quantification.
Evaluate forecasting models rigorously using industry-standard accuracy metrics including MAE, RMSE, MAPE, sMAPE, MASE, residual diagnostics, and the Diebold–Mariano statistical test.
Apply ensemble forecasting and forecast combination techniques to improve predictive accuracy and model robustness.
Develop hierarchical forecasting solutions using Bottom-Up, Top-Down, Middle-Out, and MinTrace reconciliation methods for enterprise forecasting environments.
Implement multi-step forecasting strategies using Direct, Recursive, Multi-Input Multi-Output (MIMO), Global, and Local forecasting models.
Build complete end-to-end forecasting pipelines for retail demand planning, energy load forecasting, financial time series, and multi-SKU supply chain forecasting.
Select the most appropriate forecasting methodology based on business context, data characteristics, computational complexity, interpretability, and operational requirements.
Communicate forecasting insights effectively through professional visualizations, uncertainty quantification, and decision-ready analytical reports.
Develop production-ready forecasting solutions in Python that integrate statistical methods, machine learning, Bayesian inference, and deep learning to solve real-world business forecasting problems across multiple industries.
Upon Completion, You Will Be Able To
Build end-to-end forecasting pipelines in Python from raw data to deployment-ready models.
Apply statistical, machine learning, Bayesian, and deep learning forecasting techniques with confidence.
Quantify and communicate forecast uncertainty using modern probabilistic methods.
Compare, validate, and select forecasting models using rigorous statistical evaluation techniques.
Solve complex forecasting problems involving multiple products, hierarchical structures, and multiple forecasting horizons.
Deliver enterprise-grade forecasting solutions for supply chain, retail, finance, energy, and manufacturing applications.
Apply industry best practices for forecasting model development, evaluation, and business decision support.

Reader Reviews

No reviews yet

Be the first to share your thoughts!

Related Books

Forecasting With Python Version 2
Bestseller
Forecasting

Forecasting With Python Version 2

by Krish Naidu

₹1,500