A Comprehensive Statistical and Machine Learning Study on Myntra Fashion Products

A Comprehensive Statistical and Machine Learning Study on Myntra Fashion Products

Rohan J. Kende

Department of Statistics

Dr. D. Y. Patil Arts, Commerce and Science College, Pune, India

Pranit. B. Bolli

Department of Statistics

Dr. D. Y. Patil Arts, Commerce and Science College, Pune, India

Abstract – The rapid growth of e-commerce platforms has generated large volumes of product and customer interaction data, creating opportunities for data-driven decision-making in online fashion retail. This study presents a comprehensive analytical framework for examining Myntra fashion product data using statistical methods, exploratory data analysis (EDA), natural language processing (NLP), and machine learning techniques. A dataset containing product attributes such as brand, category, price, discount percentage, ratings, reviews, and textual descriptions was cleaned and preprocessed to ensure analytical reliability. Exploratory analysis revealed strong patterns in pricing distribution, discount strategies, brand performance, color and size preferences, and rating behavior. Statistical hypothesis testing, including ANOVA, t-tests, chi- square tests, correlation, and regression analysis, confirmed significant differences in discount percentages across brands and a strong positive relationship between list

price and sale price. NLP techniques such as TF-IDF keyword extraction, text

clustering, and sentiment analysis identified dominant product themes and predominantly positive description sentiment. Multiple machine learning models, including Decision Tree, Support

Vector Machine, Random Forest, and KNN, were applied for price segment and discount-level prediction, showing high classification performance. A content-based recommendation system was also developed using category, price, ratings, and review features to generate personalized product suggestions. The results demonstrate that integrated statistical and machine learning approaches can effectively extract actionable insights, support pricing and promotion strategies, and enhance recommendation capabilities in online fashion retail environments.

Keywords

E-commerce Analytics, Myntra, Fashion Products, Exploratory Data Analysis, NLP, Sentiment Analysis, Machine Learning, Recommendation System, Statistical Testing, Pricing and Discount Analysis.

1. INTRODUCTION

   In recent years, online fashion platforms have transformed the retail industry by providing large-scale digital catalogs and personalized shopping experiences. Platforms such as Myntra host thousands of products across brands, categories, and price ranges. This generates rich datasets that can be analyzed to understand consumer preferences, pricing structures, and promotional strategies.

   This study focuses on applying statistical analysis, exploratory data analysis, natural language processing, and machine learning methods to Myntra fashion product data. The purpose is to extract meaningful insights related to pricing patterns, brand discount strategies, rating behavior, and textual product descriptions. The research also aims to build predictive and recommendation models that replicate real-world e- commerce intelligence systems.

2. OBJECTIVES

   • Identify product and brand trends

   • Analyze price and discount behavior

   • Apply hypothesis testing and regression

   • Perform NLP on product descriptions

   • Build clustering and sentiment models

   • Develop a recommendation system

3. DATA COLLECTION

   The dataset contains about 4000 Myntra product records with attributes such as brand, category, color, size, list price, sale price, discount percentage, ratings, and descriptions, supporting both statistical and NLP analysis.

4. DATA CLEANING AND PREPROCESSING

   Data cleaning included duplicate removal, missing value handling, numeric conversion, text normalization, and feature engineering such as weighted ratings, TF-IDF vectors, and price segments.

5. EXPLORATORY DATA ANALYSIS

   EDA shows major brand dominance, mid-range price concentration, black color preference, and mostly low-to- moderate discount distribution. Insert PPT graphs: brand column chart, word map, pie chart, histogram, and Pareto chart here as figures.

   Fig. 1 Brand Product Count Column Chart

   Conclusion: From above graph we can say that most product are sold by U.S. Polo , Tommy Hilfiger, Roaster.

   Fig. 2 Brand Word Map

   Conclusion: This word map shows that the project is related to all these brands

   25.16

   % 10.66

   %

   Product sold by

   colour

   13.79

   %

   10.55

   %

   WHITE

   GREEN

   PINK

   Fig. 3 Product Color Pie Chart

   Conclusion : From above graph we can say that Black coloured products are sold mostly also we can say that peoples likes black colour as compared to others.

   Fig. 4 Sale Price Histogram

   Conclusion: From above graph, we can say that most products are priced betn 1000 to 1750 and few have higher price than 2000.

   Discount

   100%

   50%

   0%

   Fig. 5 Discount Pareto Chart

   Conclusion : This graph shows that most products are have low discounts mainly between 0 to 20%. A smaller number of products fall in the higher discount categories.

   The trend line decreases sharply, which indicates that as discount increases the number of items offering that discount decreases.

6. PRICE AND DISCOUNT ANALYSIS

   Analysis shows sale price is consistently below list price, strong positive price correlation, and stable ratings even at higher discounts. Insert PPT scatter plots, line charts, heatmaps, and boxplots here.

   Fig. 6 Brand Discount Column Chart

   Conclusion : This graph shows U.S. Polo, H&M , Tommy Hillfiger gives more discount as compared to other brands.

   Fig. 7 List vs Sale Price Line Chart

   Conclusion : The graph compares the List Price and Sale Price trends across multiple produts. It clearly shows that the Sale Price is consistently lower than the List Price indicating that most products are sold at a discount.

   Fig. 8 List vs Sale Price Scatter Plot

   Conclusion : The scatter plot shows a strong positive relationship between list price and sale price. As the list price increases, the sale price also increases, but sale prices always remain lower than list prices.

   Fig. 9 Discount vs Rating Scatter Plot

   Conclusion :

   The scatter plot shows that products with moderate to high discounts (20%50%) still maintain strong average ratings, mostly between 3.0 and 5.0.

   This indicates that higher discounts positively affect customer ratings.

   Fig. 10 Correlation Heatmap

   Conclusion : Correlation heatmap confirms strong positive correlation between list and final price and negative relation with discount percent.

   Product selled by size

   L M S XL XXL

   XL

   M

   XXL

   L

   S

   Fig. 11 Size Tree Map

   Conclusion : This shows that most products of size S, M are sold as compared to other sizes.

7. NLP Analysis

TF-IDF reveals dominant keywords related to mens apparel and brands. Sentiment analysis shows mostly positive product descriptions. Insert TF-IDF and sentiment graphs from PPT here.

Fig. 12 TF-IDF Keyword Chart

Conclusion : The TF-IDF keyword chart highlights the most important and frequently appearing terms in the product descriptions.

Y = 603.87 + 1.15X

R² 0.70.

1. MACHINE LEARNING MODELS

   Decision Tree, SVM, Random Forest, KNN, and Regression models achieved high classification performance for price and discount segments.

   Fig. 15 Decision Tree Diagram

   Fig. 13 Sentiment Score Output

   Conclusion :The sentiment analysis of product descriptions shows that the majority of the kurta-related descriptions carry a positive sentiment.

   Most sentiment polarity values are above 0, indicating

   favourable wording such as stylish, versatile,

   embroidered, and printed, which contribute to a positive

   perception of the products.

   VIII. STATISTICAL TESTING

   t-test shows significant difference between list and sale price (p < 0.05).

   ANOVA shows significant difference in discount percentages across brands.

   Chi-square shows no association between brand and sub- category.

   Correlation between list and sale price 0.84.

   Regression model:

   Conclusion : The model clearly learns that product price (both list and sale price) strongly determines the price segment. Most classifications are highly pure, showing that the decision tree effectively separates budget, mid-range, and premium products using only price-related features.

2. RECOMMENDATION SYSTEM

   A content-based recommendation system using category, price, ratings, and reviews provides similar product suggestions. Insert recommendation output figure here.

   Fig. 16 Recommendation Output Table

   Conclusion : A content-based recommendation system was developed using product category, price, ratings, and reviews. The system computes similarity and suggests related products. This demonstrates how recommendation engines can be built using product metadata without user history.

3. CONCLUSION

   Statistical analysis, NLP, and machine learning methods effectively analyze fashion e-commerce data and reveal useful business insights. The study identifies strong relationships between list price, sale price, and discount strategies across major brands. Results show that higher discounts do not reduce customer ratings, indicating maintained perceived quality. NLP techniques highlight positive sentiment and strong brand emphasis in product descriptions. Machine learning models successfully predict price and discount segments, while the recommendation system provides relevant product suggestions. Overall, these approaches support better pricing decisions, targeted promotions, personalization, and improved customer experience in online fashion retail platforms

4. REFERENCE

1. Breiman, L. (2001). Random forests. Machine Learning, 45(1), 532.

2. Benesty, J., Chen, J., Huang, Y., & Cohen, I. (2009). Pearson Correlation Coefficient. Springer.

3. Cortes, C., & Vapnik, V. (1995). Support-vector networks. Machine Learning.

4. Breiman, L., Friedman, J., Stone, C. J., & Olshen, R. A. (1984). Classification and Regression Trees. CRC Press.

5. Ricci, F., Rokach, L., & Shapira, B. (2011). Recommender Systems Handbook. Springer.

6. Salton, G., & McGill, M. J. (1983). Introduction to Modern Information Retrieval. McGraw-Hill

7. MacQueen, J. (1967). Some methods for classification and analysis of multivariate observations. University of California Press.

8. Hastie, T., Tibshirani, R., & Friedman, J. (2009). The Elements of Statistical Learning. Springer.

9. Bishop, C. M. (2006). Pattern Recognition and Machine Learning. Springer.

10. Manning, C. D., Raghavan, P., & Schütze, H. (2008). Introduction to Information Retrieval. Cambridge University Press.

11. Jurafsky, D., & Martin, J. H. (2023). Speech and Language Processing. Pearson.

12. Aggarwal, C. C. (2015). Data Mining: The Textbook. Springer.

13. Aggarwal, C. C. (2016). Recommender Systems: The Textbook. Springer.

14. Liu, B. (2012). Sentiment Analysis and Opinion Mining. Morgan & Claypool.

15. Han, J., Kamber, M., & Pei, J. (2011). Data Mining: Concepts and Techniques. Morgan Kaufmann.

16. Pedregosa, F. et al. (2011). Scikit-learn: Machine Learning in Python.

    Journal of Machine Learning Research.

17. Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013). Efficient Estimation of Word Representations in Vector Space