Precise order quantity forecasting for fashion retailers is difficult, because of the specific nature of fashion products namely long lead times, seasonality, and product attributes such
as sizes, colours, and cuts. This thesis contributes to order quantity forecasting for fashion products by the use of regression analysis. For this purpose, forecasting techniques in general, and parametric as well as nonparametric regression analysis in articular are presented. This is followed by fundamentals of data mining, specifically data preprocessing and data warehousing, in order to be able to apply regression analysis on historical sales data. Furthermore, to examine the quality of forecasts a method for
evaluating the economical benefit of order quantity forecasting was developed.
As a next step, the presented methods for forecasting were applied to historical sales data. Therefore, sales data was analysed, regression models were applied and forecasts were
calculated and evaluated finally. This thesis is concluded by suggesting a forecasting implementation and by discussing the contributions to order quantity forecasting.
Table of Contents
1 Introduction
2 Problems of Fashion Retailing
2.1 Fashion Business
2.1.1 Characteristics of Fashion and Sports Equipment Products
2.1.2 Problems of Fashion Purchasing
2.1.3 Recent Developments in Fashion Purchasing
2.2 Research Questions and Goals of the Thesis
2.3 Outline of the Thesis
3 Forecasting
3.1 Fundamentals
3.2 Sales Forecasting
3.3 Sales Forecasting for Fashion Products
4 Regression Analysis
4.1 Fundamentals
4.2 Parametric Regression Analysis
4.2.1 Quality of the Estimated Regression Model
4.2.2 Nonlinear Regression Analysis
4.3 Nonparametric Regression Analysis
4.3.1 Binning and Local Averaging
4.3.2 Kernel Estimation
4.3.3 Local Polynomial Regression
4.3.4 Quality of the Estimated Nonparametric Regression Model
4.3.5 Nonparametric Multiple Regression Analysis
5 Data Mining
5.1 Fundamentals
5.1.1 Knowledge Discovery from Data
5.1.2 Measurement Scales
5.2 Preprocessing of Data
5.2.1 Data Cleaning
5.2.2 Data Transformation
5.2.2.1 Normalisation of Interval and Ratio Scaled Data
5.2.2.2 Normalisation of Ordinal Data
5.2.2.3 Normalisation of Alpha Variables
5.2.3 Data Reduction
5.3 Data Warehousing
5.3.1 Data Warehousing for Fashion Retailers
6 Economical Quality of Forecasts
6.1 Product Costing and Pricing
6.2 Costs of Overstocking and Understocking
6.3 Evaluating the Economical Quality of Forecasts
7 Application of Forecast
7.1 Calculating Regression Analysis by MATLAB
7.2 Preliminary Examination of Data
7.2.1 Data Description
7.2.2 Data Preprocessing after Export from Data Warehouse
7.2.3 Examinations of Sizes over Time
7.2.3.1 Examination of Sizes during the Season
7.2.3.2 Examinations of Sizes for the same Season over several Years
7.2.4 Examinations of Sizes over Stores
7.3 Forecast and Evaluation Process
7.4 Modelling
7.4.1 Univariate Approach
7.4.2 Multivariate Approaches
7.4.2.1 Surface Fitting Using a Parametric Polynomial Regression Model
7.4.2.2 Surface Fitting Using a Custom Equation Regression Model
7.4.2.3 Surface Fitting Using a Nonparametric Lowess Regression Model
7.5 Forecasting
7.5.1 Evaluation of the Forecast
7.5.1.1 Hypothesis 1: Actual Sales Data Represents the Demand of Products
7.5.1.2 Hypothesis 2: Actual Sales Data is Biased by Original Order
7.5.1.3 Conclusions on Hypotheses
7.6 Practical Challenges
7.7 Suggesting an Implementation
8 Conclusions
8.1 Results of the Thesis
8.1.1 General Results
8.1.2 Data Understanding
8.1.3 Data Modelling
8.1.4 Forecasting
8.1.5 Implementation
8.1.6 Summary of Results
8.2 Future Research
Research Objectives and Themes
The primary goal of this thesis is to enhance long-term order quantity forecasting for fashion products by utilizing historical sales data through regression analysis. It addresses the complexities of fashion retailing—such as seasonality, long lead times, and specific product attributes—and seeks to evaluate the economic benefits of improved forecasting methods to increase company efficiency.
- Application of regression analysis in fashion retailing
- Methodologies for data preprocessing and data warehousing
- Evaluation of forecasting accuracy and economic impact
- Mathematical modeling of size distributions in fashion retail
Auszug aus dem Buch
4.3.3 Local Polynomial Regression
Local polynomial regression is the most common nonparametric method for regression and builds the basis for nonparametric multiple regression analysis. Essentially, local polynomial regression is based on the kernel estimation and the polynomial regression model (compare Section 4.2.2). The starting point is the polynomial regression model of (4.32) in the form of Y = β0 + β1X + β2X2 + ... + βpXp, where p is the polynomial degree of the polynomial regression. The polynomial regression is used for fitting the data in the window, utilising the weights of the kernel estimator. The local kernel weights wi have the same form as for the kernel estimator wi = φ((xi - x)/h), and using the polynomial model, the following equation is used for fitting the data in the window: yi = β0 + β1(xi - x) + β2(xi - x)2 + ... + βp(xi - x)p + ei. Objective of the regression is then to minimise a weighted residual sum in the form of ∑i=1n wiei2 → min, which is also called weighted least squares regression (WLS). After the WLS is computed, the estimation ŷ at the focal point x is β0. (cf. Fox 2000a, Ch.4)
The advantage of local polynomial regression compared to kernel estimation is that the local regression reduces the boundary bias. This is because by utilising local polynomial regression, also the slope of the observations at the focal point is considered for the estimation of the regression. The more flexible the regression should be, the higher polynomials could be applied. However, the higher polynomials are applied, the greater the variation of the regression will be which could lead to overfitting. Moreover, higher polynomials require more computing capacity. For this reason, in practice local linear regression (polynomial degree p = 1) is often used, which already shows improved results compared to a kernel estimator (polynomial degree p = 0). (cf. Cleveland 1979)
Summary of Chapters
1 Introduction: Provides an overview of the challenges in fashion retailing and outlines the need for accurate demand forecasting to improve cost structures.
2 Problems of Fashion Retailing: Discusses the characteristics of fashion products, such as seasonality and long lead times, and identifies the core challenges in purchasing.
3 Forecasting: Reviews general forecasting techniques and differentiates between subjective and model-based methods, specifically for fashion contexts.
4 Regression Analysis: Details parametric and nonparametric regression models as tools for numeric prediction in data mining.
5 Data Mining: Explores fundamental data mining concepts, including data cleaning, transformation, and warehousing necessary for accurate analysis.
6 Economical Quality of Forecasts: Establishes a framework for evaluating the monetary benefit of forecasts, focusing on overstocking and understocking costs.
7 Application of Forecast: Demonstrates the practical application of regression models using MATLAB, including data preprocessing and evaluation against real sales data.
8 Conclusions: Summarizes the results of the thesis, discusses findings regarding forecast implementation, and proposes directions for future research.
Keywords
Sales Forecast, Regression Analysis, Fashion Purchasing, Data Mining, Order Quantity, Forecasting Techniques, Data Warehousing, Nonparametric Regression, Parametric Regression, Fashion Retail, Inventory Costs, Overstocking, Understocking, Decision Support System, MATLAB
Frequently Asked Questions
What is the core focus of this thesis?
The thesis focuses on improving the long-term forecasting of order quantities for seasonal fashion and sports equipment products using regression analysis techniques.
What are the primary challenges addressed in the work?
The main challenges are the specific characteristics of fashion products, including short life cycles, long lead times, seasonality, and the complexity of managing variations in sizes, colors, and cuts.
What is the primary goal of the research?
The goal is to determine the conditions under which long-term order quantity forecasts can be improved based on historical sales data, ultimately aiming to reduce inventory costs and prevent overstocking or understocking.
Which scientific methods are applied?
The author uses both parametric and nonparametric regression analysis, specifically utilizing tools like the MATLAB surface fitting tool to model sales data and calculate demand forecasts.
What does the main body of the work cover?
The main body covers the theory of forecasting, regression analysis, data mining and warehousing, the development of an economic evaluation method for forecasts, and a practical application study using real-world test data.
Which keywords best characterize this work?
Key terms include Sales Forecast, Regression Analysis, Fashion Purchasing, Data Mining, and Order Quantity.
Why are standard regression models sometimes insufficient for this data?
Standard linear models may fail due to complex, nonlinear relationships between variables; thus, the author explores nonparametric methods like Lowess to better fit the specific sales patterns of fashion items.
What is the significance of the "Hypothesis 2" in the application chapter?
Hypothesis 2 assumes that actual sales data might be biased by the original order (due to limited supply or display constraints) and suggests that a revised forecast based on sub-category demand models can better capture true consumer potential.
How does the author propose implementing these findings?
The author proposes the development of a Decision Support System (DSS) integrated into the retailer's IT landscape, where the system provides baseline forecasts that purchasers can adjust based on their expert knowledge and market trends.
- Quote paper
- Peter Hirschbichler (Author), 2010, Order Quantity Forecasting for the Fashion Industry, Munich, GRIN Verlag, https://www.hausarbeiten.de/document/164751
