Comparative Performance Analysis of Machine Learning Algorithms in the Retail Industry: Black Friday Sales Forecasting

Vahid Sinap

Makale detayları

Comparative Performance Analysis of Machine Learning Algorithms in the Retail Industry: Black Friday Sales Forecasting

Yazarlar:

Vahid Sinap

Yayıncı:

Selcuk University

Ana Başlıklar:

Satış Yönetimi | Tüketici Davranışı | Pazarlama (Diğer)

Dergi:

Selçuk Üniversitesi Sosyal Bilimler Meslek Yüksekokulu Dergisi Vol.27 Issue 1, pp.65-90

Belge Tipi:

Makale

Yayın Tarihi:

4/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001316815

pISSN:

2564-7458

Anahtar kelimeler:

Özet:

 The expansion of branch networks of large retail chains, the growth of their customer base, and the increasing diversity of customer profiles are exacerbating the complexity of sales forecasting processes. Managing this diversity and its implications presents a significant challenge for retailers in terms of both strategic planning and operational implementation. At this point, developing customer segmentation and personalized marketing strategies, determining unique approaches for each customer group, and effectively managing this diversity are becoming increasingly crucial. The emerging technologies, particularly machine learning methods, present the potential to cope with these challenges. In light of this, the main objective of the research is to perform sales forecasting on a retail companys Black Friday sales data using machine learning algorithms named Linear Regression, Random Forest Regression, K-Nearest Neighbors Regression, XGBoost Regression, Decision Tree Regression and LightGBM Regression and determine the best performing algorithm by comparing their performances. It is also aimed to tune the hyperparameters using GridSearchCV and examine the effect of these adjustments on the performance of the models. Additionally, Exploratory Data Analysis will be conducted on the dataset to create a sample example for businesses in the retail sector on how they can extract useful information from their available data and effectively evaluate it. According to the results obtained from the research, the most successful algorithm in predicting sales was the XGBoost Regression with hyperparameters tuned using GridSearchCV. It has been determined that the majority of the companys customers consist of individuals aged 26-35, with male customers making significantly higher purchases compared to females and single customers spending more than married ones. Furthermore, when examining the average amount of purchases made by each age group, it was identified that those within the range of 51-55 years had the highest average spending rate.

 The expansion of branch networks of large retail chains, the growth of their customer base, and the increasing diversity of customer profiles are exacerbating the complexity of sales forecasting processes. Managing this diversity and its implications presents a significant challenge for retailers in terms of both strategic planning and operational implementation. At this point, developing customer segmentation and personalized marketing strategies, determining unique approaches for each customer group, and effectively managing this diversity are becoming increasingly crucial. The emerging technologies, particularly machine learning methods, present the potential to cope with these challenges. In light of this, the main objective of the research is to perform sales forecasting on a retail companys Black Friday sales data using machine learning algorithms named Linear Regression, Random Forest Regression, K-Nearest Neighbors Regression, XGBoost Regression, Decision Tree Regression and LightGBM Regression and determine the best performing algorithm by comparing their performances. It is also aimed to tune the hyperparameters using GridSearchCV and examine the effect of these adjustments on the performance of the models. Additionally, Exploratory Data Analysis will be conducted on the dataset to create a sample example for businesses in the retail sector on how they can extract useful information from their available data and effectively evaluate it. According to the results obtained from the research, the most successful algorithm in predicting sales was the XGBoost Regression with hyperparameters tuned using GridSearchCV. It has been determined that the majority of the companys customers consist of individuals aged 26-35, with male customers making significantly higher purchases compared to females and single customers spending more than married ones. Furthermore, when examining the average amount of purchases made by each age group, it was identified that those within the range of 51-55 years had the highest average spending rate.

 The expansion of branch networks of large retail chains, the growth of their customer base, and the increasing diversity of customer profiles are exacerbating the complexity of sales forecasting processes. Managing this diversity and its implications presents a significant challenge for retailers in terms of both strategic planning and operational implementation. At this point, developing customer segmentation and personalized marketing strategies, determining unique approaches for each customer group, and effectively managing this diversity are becoming increasingly crucial. The emerging technologies, particularly machine learning methods, present the potential to cope with these challenges. In light of this, the main objective of the research is to perform sales forecasting on a retail companys Black Friday sales data using machine learning algorithms named Linear Regression, Random Forest Regression, K-Nearest Neighbors Regression, XGBoost Regression, Decision Tree Regression and LightGBM Regression and determine the best performing algorithm by comparing their performances. It is also aimed to tune the hyperparameters using GridSearchCV and examine the effect of these adjustments on the performance of the models. Additionally, Exploratory Data Analysis will be conducted on the dataset to create a sample example for businesses in the retail sector on how they can extract useful information from their available data and effectively evaluate it. According to the results obtained from the research, the most successful algorithm in predicting sales was the XGBoost Regression with hyperparameters tuned using GridSearchCV. It has been determined that the majority of the companys customers consist of individuals aged 26-35, with male customers making significantly higher purchases compared to females and single customers spending more than married ones. Furthermore, when examining the average amount of purchases made by each age group, it was identified that those within the range of 51-55 years had the highest average spending rate.

 Büyük perakende zincirlerinin şube ağlarının genişlemesi, müşteri tabanlarının büyümesi ve artan müşteri profili heterojenliği satış tahminleme süreçlerinin karmaşıklığını artırmaktadır. Müşteri çeşitliliği ve bu çeşitliliğin yönetilmesi, perakendeciler için hem stratejik planlama hem de operasyonel düzeyde uygulama açısından önemli bir güçlük oluşturmaktadır. Bu noktada, müşteri segmentasyonu ve kişiselleştirilmiş pazarlama stratejileri geliştirmek, her bir müşteri grubuna özel yaklaşımlar belirlemek ve bu çeşitliliği anlayarak etkili bir şekilde yönetmek önem kazanmaktadır. Gelişen teknolojiler, özellikle makine öğrenmesi yöntemleri söz konusu zorluklarla başa çıkma potansiyeli sunmaktadır. Bu kapsamda araştırmanın amacı, bir perakende firmasının Black Friday günündeki satış veri seti üzerinde Doğrusal Regresyon, Rastgele Orman Regresyonu, K-En Yakın Komşu Regresyonu, XGBoost Regresyonu, Karar Ağacı Regresyonu ve LGBM Regresyonu isimli makine öğrenmesi algoritmaları aracılığıyla satış tahminlemesi gerçekleştirmek ve algoritmaların performanslarını karşılaştırarak en iyi performans gösteren algoritmayı belirlemektir. Ayrıca, GridSearchCV kullanarak hiperparametrelerin ayarlanması ve bu ayarlamaların modellerin performanslarına etkisinin incelenmesi amaçlanmaktadır. Buna ek olarak, veri seti üzerinde Keşifsel Veri Analizleri yürütülerek, perakende sektöründeki işletmelerin ellerinde bulunan verilerden ne tür bilgiler çıkarabileceklerine ve bu bilgileri nasıl değerlendirebileceklerine ilişkin bir örnek oluşturmak araştırmanın diğer önemli bir amacıdır. Araştırmadan elde edilen sonuçlara göre, satışları tahminlemede en başarılı algoritma GridSearchCV ile hiperparametreleri ayarlanmış XGBoost Regresyonu olmuştur. Firma müşterilerinin en çok 26-35 yaş aralığında bireylerden oluştuğu, erkek müşterilerin kadınlara, bekar müşterilerin evlilere göre önemli ölçüde daha yüksek tutarlı alışverişler yaptığı saptanmıştır. Ayrıca, satın alım tutarı ortalaması bağlamında bakıldığında en yüksek harcama ortalamasına sahip yaş grubu 51-55 yaş aralığı olarak tespit edilmiştir.

Makale detayları

Comparative Performance Analysis of Machine Learning Algorithms in the Retail Industry: Black Friday Sales Forecasting

Yazarlar:

Vahid Sinap

Yayıncı:

Selcuk University

Ana Başlıklar:

Satış Yönetimi | Tüketici Davranışı | Pazarlama (Diğer)

Dergi:

Selçuk Üniversitesi Sosyal Bilimler Meslek Yüksekokulu Dergisi Vol.27 Issue 1, pp.65-90

Belge Tipi:

Makale

Yayın Tarihi:

4/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001316815

pISSN:

2564-7458

Sales Forecasting

Machine Learning

Regression Algorithms

Black Friday

Retail IndustrySatış Tahminlemesi

Makine Öğrenmesi

Regresyon Algoritmaları

Black Friday

Perakende Sektörü

Sayı

Makale detayları

Yazarlar: Vahid Sinap

Yayıncı: Selcuk University

Ana Başlıklar: Satış Yönetimi | Tüketici Davranışı | Pazarlama (Diğer)

Dergi: Selçuk Üniversitesi Sosyal Bilimler Meslek Yüksekokulu Dergisi Vol.27 Issue 1, pp.65-90

Belge Tipi: Makale

Yayın Tarihi: 4/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001316815

pISSN: 2564-7458

Sayı

Yazarlar:

Vahid Sinap

Yayıncı:

Selcuk University

Ana Başlıklar:

Satış Yönetimi | Tüketici Davranışı | Pazarlama (Diğer)

Dergi:

Selçuk Üniversitesi Sosyal Bilimler Meslek Yüksekokulu Dergisi Vol.27 Issue 1, pp.65-90

Belge Tipi:

Makale

Yayın Tarihi:

4/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001316815

pISSN:

2564-7458