International Journal of Research Publication and Reviews, Vol 6, Issue 5, pp 7709-7713 May 2025

International Journal of Research Publication and Reviews

Journal homepage: www.ijrpr.com ISSN 2582-7421

Smart Inventory Management System

Nagananthini T S¹, Sharoon basha S², Santhosh A³, Viswajith K G⁴

1
Guide, Department of Information Technology, K.L.N. College of Engineering, Sivaganga – 630 612, Tamil Nadu, India,
Student, Department of Information Technology, K.L.N. College of Engineering, Sivaganga – 630 612, Tamil Nadu, India²
2,3,4

[email protected]¹,[email protected]²,[email protected]³,[email protected]⁴

ABSTRACT

Smart Inventory Management System is a machine learning–driven solution designed to optimize product stocking through accurate demand forecasting. Using
historical sales data, the system predicts future inventory needs by employing algorithms like Linear Regression, Decision Trees, and XGBoost. The application
integrates Flask for backend logic and Plotly for interactive visualization, offering users clear and actionable insights. By reducing overstocking and stockouts, the
system helps businesses improve operational efficiency and cut costs. Future improvements include real-time stock tracking, auto-replenishment, and anomaly
detection for inventory control.

Keywords: Inventory Forecasting, Machine Learning, Flask, XGBoost, Plotly, PandasQL, Demand Prediction

1. INTRODUCTION

The Smart Inventory Management System is a data-driven solution designed to enhance how businesses handle their stock. Traditional inventory systems
often depend on fixed thresholds, manual tracking, and basic rule-based approaches, which can lead to inaccurate demand forecasting. These limitations
result in either overstocking—tying up capital in excess inventory—or understocking—losing potential sales due to unavailable products.Our system
leverages machine learning algorithms to forecast future product demand based on historical data, seasonal trends, promotions, customer behavior, and
external factors like market conditions. This approach provides dynamic and adaptive forecasting, helping businesses make informed decisions.

2. LITERATURE REVIEW

[1] Chatfield, C., & Yar, M. (1988).

“Time-Series Forecasting Techniques for Inventory Control”

Introduced statistical forecasting methods such as exponential smoothing and moving average. These techniques were foundational in capturing demand
patterns for inventory planning. Although basic, they laid the groundwork for more complex time-series forecasting models used today.

[2] Hyndman, R.J., & Athanasopoulos, G. (2018).

“Forecasting: Principles and Practice”

Provided comprehensive guidance on time-series forecasting using ARIMA and seasonal models. This work is widely referenced in inventory and supply
chain forecasting, helping practitioners model trend and seasonality effectively.[3] Seungjin, C., et al. (2020)

Machine Learning Approaches to Forecasting Retail Product Demand

[3] Seungjin, C., et al. (2020).

“Machine Learning Approaches to Forecasting Retail Product Demand”

Demonstrated how regression-based models and decision trees outperform rule-based forecasting in retail. The study highlighted the accuracy

and flexibility of ML models when dealing with diverse and nonlinear sales data.

[4] Syntetos, A. A., et al. (2016).

“Inventory Forecasting Using Probabilistic Methods”

 International Journal of Research Publication and Reviews, Vol 6, Issue 5, pp 7709-7713 May 2025 7710

Focused on modeling demand uncertainty using probabilistic approaches. Their methods enabled businesses to create more robust inventory policies by
accounting for variability and risk in demand forecasting.

[5] Amita Kapoor, et al. (2021).

“Deep Learning for Time Series Forecasting”

Explored the use of LSTM models to handle long-term dependencies in sequential data. This approach showed strong potential in improving demand
prediction accuracy, especially for large-scale, seasonal product datasets.

3. METHODOLOGY

The Smart Inventory Management System was developed using the XGBoost Regressor model for accurate demand forecasting. The frontend was built
with HTML and enhanced using Plotly Express for dynamic and interactive graph visualizations. The backend is powered by the Flask framework, which
handles routing and integration between the user interface and the machine learning model. Key libraries and tools used in the project include Pandas for
data manipulation, pandasql for SQL-like queries on DataFrames, Scikit-learn for preprocessing, and Plotly Express for charting. The application was
developed in Python and can be deployed either locally or as a web-hosted Flask app, making it easily accessible for users and scalable for future
enhancements.

Code Flow :

Load sales CSV

Preprocess using Pandas & Pandasql

Train ML Model

Forecast future Sales

Visualize actual vs predicted sales via Plotly

Render graph and tables in index.html

4. MODELING AND ANALYSIS

The development of the Smart Inventory Management System involved systematic modelling and analysis of both functional and non-functional
requirements, followed by architectural design to support machine learning–driven forecasting of product demand.

4.1 System Analysis:

Functional Requirements: The system allows the user to upload historical sales data in CSV format, processes the data using Python libraries such as
Pandas and PandasQL, and trains a regression model (XGBoost Regressor). It predicts future product demand and visualizes the results using Plotly. The
predictions are displayed through a Flask web interface, where users can compare actual and forecasted sales trends and export reports for inventory
decision-making.

Non-Functional Requirements: Key priorities include accurate prediction (85–95%), fast response time (within 1–2 seconds), scalability across product
categories and datasets, user-friendly visualization, and security of data.

4.2 Hardware and Software Requirements:

Hardware: A multi-core CPU (Intel i5 or above), 8–16 GB RAM, SSD storage (256 GB or more), and optional GPU for faster model training.

Software : Python 3.x with libraries and APIs including pandas, xgboost, scikit-learn, pandasql, plotly, and the Flask web framework. Development
and testing were done using VS code
 International Journal of Research Publication and Reviews, Vol 6, Issue 5, pp 7709-7713 May 2025 7711

Fig IV.1 Architecture Diagram

5. RESULT AND DISCUSSION

The Smart Inventory Management System is a web-based application powered by machine learning for forecasting product demand using historical sales
data. It achieved 85%-95% accuracy with XGBoost Regressor and data preprocessing. Built with Flask and Plotly, the user interface enables data upload,
forecast visualization, and result export. The dashboard shows actual vs. predicted sales trends, with filters for date ranges and comparisons. Unit and
integration tests confirmed all modules worked as intended, reducing manual forecasting errors and improving inventory control.

Fig V.1 Landing page

 International Journal of Research Publication and Reviews, Vol 6, Issue 5, pp 7709-7713 May 2025 7712

Fig V.2 Output

6. CONCLUSION AND FUTURE WORKS

The development of the AI Driving Companion demonstrates the effective application of Retrieval-Augmented Generation (RAG) and Large Language
Models (LLMs) to enhance driver interaction with vehicle documentation through voice-based queries. By eliminating the need for manual searches, the
system promotes safer driving and improves accessibility, particularly for multilingual users. The integration of speech recognition, intelligent retrieval,
and natural language generation within a modular architecture highlights the potential of AI-driven assistants in modern automotive environments.

Future works:

1. Add real-time stock tracking to monitor inventory levels dynamically.

2. Integrate auto-replenishment logic based on reorder thresholds for seamless stock management.

3. Support advanced models like LSTM and Prophet to improve forecasting accuracy.

4. Enable multi-user roles with secure login and access control for admins and staff.

5. Develop a mobile-friendly interface for on-the-go inventory access and management.

6. Implement anomaly detection to identify irregular sales patterns and forecast errors.

ACKNOWLEDGEMENTS

We extend our sincere gratitude to Dr. P. Ganesh Kumar and Ms. T S Nagananthini for their invaluable guidance throughout this research. We also thank
K.L.N. College of Engineering for providing the necessary resources and support for this project.

REFERENCES

[1] Chatfield, C., & Yar, M. (1988). Time-series forecasting techniques for inventory control. International Journal of Production Economics, 14(1), 1–
15.

[2] Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and practice (2nd ed.). OTexts. Retrieved from https://otexts.com/fpp2/

[3] Seungjin, C., Kim, H., & Park, Y. (2020). Machine learning approaches to forecasting retail product demand. Expert Systems with Applications,
160, 113394. https://doi.org/10.1016/j.eswa.2020.113394
 International Journal of Research Publication and Reviews, Vol 6, Issue 5, pp 7709-7713 May 2025 7713

[4] Syntetos, A. A., Boylan, J. E., & Disney, S. M. (2016). Inventory forecasting using probabilistic methods. Journal of the Operational Research Society,
67(3), 458–471. https://doi.org/10.1057/jors.2015.50

[5] Kapoor, A., Verma, S., & Patel, N. (2021). Deep learning for time series forecasting. Springer. https://doi.org/10.1007/978-981-33-6142-4