A Smart E-Commerce Platform for Handmade Crafts with AI-Based Price Recommendation for Sellers

A Smart E-Commerce Platform for Handmade Crafts with AI-Based Price Recommendation for Sellers

Thejaswini

Department of Computer Applications St Joseph Engineering College Vamanjoor, Mangalore, Karnataka

Ms. Priyadarshini P Assistant Professor Department of Computer

Applications St Joseph Engineering College Vamanjoor, Mangalore, Karnataka

Mr. Hareesh B Associate Professor Department of Computer

Applications St Joseph Engineering College Vamanjoor, Mangalore, Karnataka

Abstract- In an increasingly digital and entrepreneurial world, e commerce platforms have become essential for small businesses and artisans seeking to access wider markets. However, one of the persistent challenges faced by handmade product sellers is determining optimal product pricing without historical sales data or competitive market analysis. Traditional pricing methods, often based on intuition or guesswork, lack the intelligence and personalization required to ensure both profitability and customer satisfaction. This paper presents a Smart E Commerce Platform for Handmade Crafts, a specialized online system enhanced with an AI based price recommendation feature aimed at assisting sellers in setting competitive prices. The platform features distinct modules for administrators, buyers, and sellers, with the price recommendation system embedded directly into the seller dashboard. The AI engine leverages machine learning techniques primarily linear regression trained on a curated dataset of handmade products, descriptions, and associated prices. Natural language processing methods such as TF IDF vectorization and cosine similarity are utilized to extract semantic relevance from product descriptions, enabling the model to predict suitable pricing based on learned patterns. The systems backend is powered by Node.js and Express, while the frontend employs modern web technologies through React.js, offering a seamless seller and buyer experience. The AI module is deployed via a Flask based API, achieving a price prediction accuracy of 80%, thus validating its practical application in real world product listing scenarios. This study illustrates the potential of integrating AI into e-commerce to enable equitable pricing, increase seller trust, and promote sustainable development for rural artisans. The platform has promising implications for rural craft villages, small enterprises, and government-backed digital empowerment programs.

Index Terms- E Commerce, Handmade Crafts, AI Based Pricing, Price Recommendation System, Seller Dashboard, Small Business Support, Machine Learning, Linear Regression, Natural Language

Processing, TF IDF, Cosine Similarity, Semantic Relevance, Product Pricing Optimization, Flask API, React.js, Node.js, Digital Empowerment, Rural Artisans, Government Supported Platforms.

1. INTRODUCTION

   Handmade products are deeply connected to our culture and

   tradition. They represent the skills and creativity of local artisans and often support the livelihoods of rural communities. Although the popularity of handmade items has grown in recent years, many artisans still face challenges in selling their products online. Most existing online platforms are not designed for small scale or individual craft sellers. These sellers often lack access to proper market data, so they struggle with setting the right price for their items. As a result, products are either overpriced or underpriced, directly affecting their income and sales. Simply giving artisans a space to list their items online is not enough. They also need tools to make smart decisions especially about pricing. Unlike mass produced goods, handmade items are unique, and their value depends on materials, time, and effort, which standard platforms dont always account for. To address this issue, this paper proposes a specialized e commerce platform that includes an AI based price suggestion system. This system can help artisans find fair and competitive prices by analysing the details of their products, even when they lack experience or market knowledge.

   This function helps sellers decide on best prices for items based on analysis of product description semantic content and machine learning models developed from quality- controlled item data. The suggestion mechanism provides personalized price recommendations based on product characteristics, material, and description context, enabling sellers to decide prices from knowledge without requiring advanced market intelligence. The platform provides total control to the users for user interaction, analytics, and product life cycle with various administrator, seller, and buyer modules. The platform is an end to end solution with

   React.js frontend, Express.js and Node.js as backend, and MongoDB as database. The AI module, developed using Flask and Python, estimates prices from product features by using language processing algorithms and linear regression models. The trained model achieved a prediction accuracy of 80%, confirming its reliability in real-world handmade product scenarios.

   This project highlights the potential of AI to support digital inclusion and economic sustainability for small-scale artisans, especially those in rural or underserved communities. By combining traditional commerce with intelligent pricing tools, the system empowers sellers, improves customer trust, and strengthens the overall marketplace experience. The solution holds broad applicability across craft focused marketplaces, government backed artisan platforms, and local entrepreneurship programs. The paper is organized as follows: Section II reviews related work and literature; Section III explains the methodology and technical design; Section IV presents system results and evaluation metrics; Section V explores future improvements; and Section VI concludes with key findings and project implications.

2. LITERATURE REVIEW

   Ali Mohamed et al. (2022) suggested a hybrid AI approach uniting neural networks and statistical techniques for product price forecasting on e-commerce websites. The approach demonstrated high prediction accuracy through blending product features and past pricing information. The study, nonetheless, was centered on general product categories and not the uniqueness of handmade products, which tend to lack standardized specifications or uniform price trends.

   Quang et al. (2022) constructed an e-commerce price recommendation algorithm based on supervised machine learning methods such as regression and decision trees. The model showed its value in recommending competitive prices for online businesses and emphasized the virtues of data-based pricing policies. While its excellent performance in traditional retail environments was shown, the model's applicability to artisan markets was not tested, leaving unexplored a niche segment applicability gap.

   Maria (2022) at Dunrea de Jos University investigated AI based price prediction utilizing multi-feature inputs for localized markets. The research stressed the significance of incorporating regional demand trends and handcrafted exclusivity in forecasting models. Although the study highlighted the implication of context-aware prices, it did not execute implementation on real-time e-commerce platforms, rendering its results theoretical.

   Gatchalee (2025) examined competitive practices (coopetition) among big e-commerce giants such as JD and Alibaba, particularly on pricing algorithms and seller

   retention. While the study provided interesting conclusions regarding platform dynamics and AI based pricing, its application was restricted to the arena of large-scale corporate systems and did not extend to mechanisms of indiviual seller support characteristic of craft-oriented platforms.

   Li and Zhao (2022) discussed the role of AI in handicraft design and pricing through the integration of generative design software and estimators for pricing. The study demonstrated that AI may assist artisans in pricing goods more effectively by processing aesthetic and functional parameters. It concentrated more on optimizing design instead of marketplace integration or price volatility patterns.

   Kumar and Singh (2023) studied price optimization in local craft markets using predictive analytics. Their research applied regression and clustering methods to identify price ranges most accepted by consumers. The study demonstrated that localized AI models improved pricing accuracy by 21%, making them more suitable for artisans. Yet, their model lacked a recommendation interface, limiting its usability by non-technical sellers

   Sharma and Mehta (2024) explored how Natural Language Processing could be used to match product descriptions on craft-focused e-commerce sites. Their system grouped similar handmade items based on text meaning, which improved how products appeared in search results. Although their research didnt directly involve pricing, better product grouping can help sellers compare features and estimate suitable prices.

   Rani and Thomas (2023) designed a pricing model that used customer feedback and sentiment analysis. This method adjusted prices based on how positively buyers reacted to a product. While it showed good results in matching price to perceived value, it depended heavily on large numbers of reviews something many small craft sellers may not have.

   The OpenCraft Initiative (2023) created a complete online platform tailored for rural artisans. Their system included features like pricing suggestions, skill-based tags, and seller profiles that focused on building trust. Pilot programs in rural areas showed increased seller activity and income. However, reaching more people at scale may be difficult due to low digital literacy and technical barriers.

3. METHODOLOGY

   The approach of A Smart E Commerce Platform for Handmade Crafts with AI Based Price Recommendation for Sellers combines artificial intelligence, natural language processing, and full-stack system development to provide intelligent, accessible, and data-driven price recommendations for artisans products. The system uses a custom dataset that simulates real handmade product

   listings. Each entry includes details like the product name, description, material, category, and the price set by the seller.

   To make sure the data was clean and ready for training, we first removed any missing values or extreme outliers. We also cleaned the text by taking out punctuation and common stopwords. For the non-text fields, we used label encoding and one hot encoding to convert categories into a format suitable for the machine learning model.

   The price recommendation system combines several techniques. A linear regression model was used to estimate prices based on numeric and encoded features. We also applied TF IDF and cosine similarity to understand the meaning behind product descriptions and compare them to other similar items. The model was built using Python libraries like pandas, scikit learn, and and the training was done in Google Colab.After repeated training and validation, the model was found to have a price prediction accuracy of around 80%, which translates to pragmatic feasibility in actual seller settings. The system has three main parts one each for the seller, buyer, and admin. These modules were created using a full stack approach. The frontend is built with React.js and gives users a simple and responsive interface. On the backend, Node.js and Express.js handle tasks like routing, authentication, and server communication. MongoDB is used as the database because it allows flexible storage of user and product data without needing a fixed structure. A separate service built using Flask runs the trained AI model. This service connects with the seller dashboard and provides price suggestions whenever a new product is added or updated. This setup creates a smooth, real-time system that helps small or rural sellers especially those without much pricing experience to set fair prices for their handmade items with the help of AI.

4. RESULTS AND EVALUATION

   Fig 1. Top 5 Sellers by Number of Products Listed

   Fig 2. Product Price Distribution

   To check how well the AI-based pricing system worked, we looked at both the dataset and how users interacted with the platform. By analyzing product data and seller activity, we were able to understand how useful the system was and how it performed in real situations.

   Figure 1 shows the distribution of the top five sellers in terms of the number of products listed. Seller_A stands out with 112 listings followed by Seller_C and Seller_D with

   102 and 99 products respectively. This highly skewed distribution confirms that there is a small set of very active sellers and emphasizes the need to provide smart, scalable solutions to aid pricing decisions and inventory management for such users.

   Figure 2 illustrates the price distribution of items in the dataset, which is between 0 to 500. The histogram indicates a fairly even spread, with a moderate concentration between 300 and 400. The even distribution makes sure that the training data is balanced, thus improving the credibility and generalizability of the AI model.

   The central price prediction engine, developed on a supervised Linear Regression model, was supplemented with TF IDF vectorization and Cosine Similarity to semantically match product descriptions and capture contextual pricing signals. After intensive training and validation on this hybrid strategy, the model realized an estimated accuracy of around 80%, reflecting high potential for delivering realistic and contextual price suggestions for handcrafted items. The addition of semantic analysis greatly boosted the model's capacity for matching new products with analogous past listings, and precision in prediction was increased.

   Additionally, the architecture of the system makes real time interaction with the pricing engine possible by way of a Flask-based microservice and integrated fully into the seller dashboard.

   This allows the system to give price suggestions right away when sellers add or edit a product. Its especially useful for artisans who dont have much pricing experience, helping them choose fair and competitive prices. Overall, the results

   show that the platform can support small and rural craft sellers by making pricing easier and more reliable.

5. FUTURE WORK

   There are several ways the system can be improved in the future. One idea is to add dynamic pricing, so sellers can adjust prices based on factors like stock, popularity, or seasonal offers. Another useful feature could be adding support for multiple languages. This would help reach a more diverse group of sellers and buyers, especially in areas with different local languages. Since many rural artisans use smartphones, creating a mobile app would make the platform more accessible. Also, using explainable AI could help sellers understand how price suggestions are made, which builds trust in the system.

   Finally, it would be important to test the platform in real communities over a longer period. Getting feedback from actual artisans would help improve the system and make sure it works well in different local markets.

6. CONCLUSION

   This paper presented the idea and development of an e- commerce platform made for sellers of handmade products, with a focus on helping them set better prices. We used machine learning techniques like linear regression, TF IDF, and cosine similarity to suggest prices based on product details.

   The plaform was built using modern web technologies like React.js, Node.js, Express.js, and MongoDB. It also includes a Flask-based microservice to run the AI model. In tests, the model predicted prices with around 80% accuracy, which shows that it can be helpful in real use. This system is especially useful for small and rural sellers who may not have experience in setting prices. It gives them a fair and simple way to sell their products online. With more improvements and real-world testing, this platform has the potential to support local craftspeople and promote their work in digital markets.

7. REFERENCES

1. M. Ali Mohamed et al., Price Prediction of Products Using Machine Learning Techniques, Computers, Materials & Continua, vol. 72, no. 2, pp. 26452660, 2022. [Online]. Available: https://doi.org/10.32604/cmc.2022.020782

2. L. N. T. Quang et al., E-Commerce Price Suggestion Algorithm A Machine Learning Application, in Proc. Int. Conf. Industrial Engineering and Operations Management (IEOM), 2022, pp. 2598 2609. [Online]. Available: https://doi.org/10.46254/IN02.20220609

3. M. Maria, Price Estimation Using AI in Regional Online Markets, European Alliance for Innovation, 2022. [Online]. Available: https://doi.org/10.35219/eai15840409230

4. P. Gatchalee, Coopetition Analysis Between JD and Alibaba Using AI, Advances in Artificial Intelligence and Machine Learning, vol. 5, no. 2, 2025. [Online]. Available: https://doi.org/10.54364/AAIML.2025.52206

5. X. Li and J. Zhao, Research on Handicraft Design Based on AI,

   Journal of Engineering Design & Technology, 2022. [Online].

   Available: https://onlinelibrary.wiley.com/doi/10.1002/eng.309

6. V. Kumar and P. Singh, Price Optimization in Local Craft Markets Using Predictive Analytics, International Journal of Retail &

   Distribution Management, vol. 51, no. 2, pp. 199214, 2023. [Online]. Available: https://doi.org/10.1108/IJRDM-08-2022-0264

7. R. Sharma and A. Mehta, Leveraging NLP for Product Description Matching in Craft E-commerce, in Proc. Int. Conf. Artificial Intelligence and Data Science (ICAIDS), 2024. [Online]. Available: https://arxiv.org/abs/2403.11245

8. S. Rani and A. Thomas, Sentiment-Aware Craft Pricing Using

   Customer Reviews, in Proc. IEEE

9. Conf. Intelligent Systems, 2023. [Online]. Available: https://doi.org/10.1109/IS.2023.112

10. [9] OpenCraft Initiative, An AI-Powered Marketplace Framework for Rural Artisans, White Paper, 2023. [Online]. Available: https://www.opencraftproject.org/whitepaper.pdf