A Smart Shopping Assistant with ML Based Price Prediction, Alerts and Deal Predictions

Abstract

Prediction is an advanced and intelligent system created to help online customers make informed purchases and save money. The core purpose of this system is to predict future product prices, seasonal discounts, and potential deals using machine learning techniques to provide customers the timely information needed to buy at the right time. The system uses advanced model systems such as Long Short-Term Memory (LSTM) networks, that are very effective in the analysis of the previous price patterns and ability to predict or forecast future patterns more accurately.The system collects real-time data using the BeautifulSoup and Requests libraries for web scraping in order to gather data, including product name, price, discounts, and availability, from various e-commerce selling platforms. The collected real-time data is then stored and managed into the MongoDB system. MongoDB is a NoSQL database, which allows for flexible data and effective management of large amounts of ongoing real-time data updates. Automated personalized recommendations and smart alerts provided by the assistant is built on the user\'s preferences, browsing history, and purchasing history, to make the user aware of changing prices, current deals, or the best time to make a purchase.To further add to the shopping experience, the system includes predictive analytics forms of automation and recommendations. Using predictive analytics, the system learns from user interaction with the site and the market.

Introduction

The Smart Shopping Assistant is a machine learning-based system designed to enhance online shopping by predicting product prices, alerts, and deals. With the rise of e-commerce platforms like Amazon and Flipkart, consumers face difficulty in determining the optimal time to purchase due to fluctuating prices influenced by demand, promotions, and competitor pricing. Existing services only provide current price listings, lacking predictive capabilities.

The proposed system addresses this by:

1. Data Collection: Automatically scraping real-time product data, including prices, ratings, reviews, and stock status, using tools like BeautifulSoup, Requests, and Selenium. The data is cleaned, standardized, and stored in databases like MongoDB for analysis.

2. Predictive Modeling: Uses LSTM (Long Short-Term Memory) networks to forecast future prices, capturing short-term fluctuations and long-term trends. Ensemble models combining ARIMA and Prophet enhance prediction accuracy.

3. Recommendation and Alerts: Provides personalized product recommendations using content-based filtering (TF-IDF, cosine similarity) and sends real-time notifications via email, SMS, or in-app alerts when predicted prices drop or deals occur.

4. Dashboard & Visualization: An interactive interface built with Matplotlib and Plotly displays live prices, trends, and recommended deals, helping users make informed purchasing decisions.

5. Continuous Learning: Models are regularly retrained with updated data to maintain prediction accuracy as market conditions change.

The system architecture follows a microservices and layered approach, ensuring scalability, modularity, and resilience. Components include data ingestion, preprocessing, prediction, recommendation, alert generation, and presentation layers. Automation and predictive analytics reduce manual monitoring while providing users with timely insights, enabling proactive and cost-effective shopping.

Key Benefits:

• Predicts price drops and future discounts

• Offers personalized recommendations

• Sends real-time alerts for optimal purchasing opportunities

• Supports scalability, efficiency, and continuous learning for accurate predictions

Conclusion

The recommended discount forecasting solution successfully combines LSTM-based forecasting, MongoDB data management, and Streamlit visualization to provide actionable insights and reliable recommendations for pricing strategies. The solution takes advantage of LSTM\'s sequential learning capabilities to forecast discounted prices based on historical sale sequences, utilizing trends that traditional models tend to overlook. MongoDB allows for a scalable and flexible data architecture for querying, storing and preprocessing large datasets on sales history. The Streamlit application brings the complexities of the predictive analytics into a friendly, interactive dashboard with Best Deal and Alert modules that can also notify users when to take action and assist in identifying the discounted price of an item. As a whole, the system is functional, scalable and can be practically applied to live retail and e-commerce environments. The results show that predictive modeling combined with dynamic visualization and alerting mechanisms assists in making decisions for businesses and customers. Unlike theoretical implementations or improvements in technology- all which were demonstrated in the project- taking advantage of both machine learning and modern database and dashboard technology can lead to intelligent data-driven systems. Future enhancements could include additional features such as customer analysis, competitor pricing data, and streaming live updates or data to enhance prediction accuracy

References

[1] Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735 [2] Brownlee, J. (2018). Deep Learning for Time Series Forecasting: Predict the Future with MLPs, CNNs and LSTMs in Python. Machine Learning Mastery. [3] Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press. [4] Chodorow, K. (2013). MongoDB: The Definitive Guide (2nd ed.). O\'Reilly Media. [5] Streamlit Inc. (2023). Streamlit Documentation - Create Interactive Web Apps for Machine Learning and Data Science. https://docs.streamlit.io [6] Brownlee, J. (2020). Time Series Forecasting With LSTMs in Python. Machine Learning Mastery. https://machinelearningmastery.com [7] Raschka, S., & Mirjalili, V. (2019). Python Machine Learning (3rd ed.). Packt Publishing. [8] Graves, A. (2012). Supervised Sequence Labelling with Recurrent Neural Networks. Springer. [9] Kingma, D. P., & Ba, J. (2015). Adam: A Method for Stochastic Optimization. In ICLR 2015. [10] Zhang, G., Eddy Patuwo, B., & Hu, M. Y. (1998). Forecasting with artificial neural networks: The state of the art. International Journal of Forecasting, 14(1), 35-62. [11] Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2018). Forecasting methods for statistical and machine learning: Issues and recommendations. PLoS ONE, 13(3), e0194889. https://doi.org/10.1371/journal.pone.0194889 [12] Zhang, C., & Lu, Y. (2021). Artificial Intelligence in E-commerce: A survey and manager implications for pricing prediction and recommendation systems. Electronic Commerce Research and Applications, 48, 101073. https://doi.org/10.1016/j.elerap.2021.101073 [13] Hyndman, R. J., & Athanasopoulos, G. (2021). Forecasting: Principles and practice (3rd ed.). OTexts. https://otexts.com/fpp3/ [14] Sun, P., Ma, X., & Zhao, Y. (2020). Short-term E-commerce sales forecasting using a hybrid deep-learning model. IEEE Access, 8, 155850– 155860. https://doi.org/10.1109/ACCESS.2020.3019355 [15] Liu, Y., Singh, P., & Sidorov, K. (2022). Intelligent price prediction and dynamic deal alert generation through hybrid machine learning approaches. Expert Systems with Applications, 198, 116870. https://doi.org/10.1016/j.eswa.2022.116870

Copyright

Copyright © 2025 P. Keerthika, K. Padma, T.T.R. Ranganayagi, G. Rajeswari. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.