INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025
www.ijltemas.in Page 397
Predictive Analytics for Inventory Optimization in Manufacturing
B R S Mendis
General Sir John Kotelawala Defense University
DOI: https://doi.org/10.51583/IJLTEMAS.2025.1408000048
Received: 30 July 2025; Accepted: 06 Aug 2025; Published: 02 September
Abstract—Accurate demand forecasting is important in the biscuit manufacturing business to optimize inventory management,
maintain operational efficiency and reduce costs. Currently, the scenario subjected company has produced over 50 products under
different brand names as well as SKUs. To get demand forecasting, they rely on the manual methods conducted by the sales and
administration and marketing departments. These manual methods are performed using tools like excel, they are prone to errors,
data duplication, and inefficiencies which are leading to unreliable forecasts and dependencies on individuals. To address these
challenges, this project aims to develop predictive analytics forecasting models. To build this model, machine learning techniques
will be employed such as Random Forest, Time series and Regression which will be used to analyze historical sales data to identify
the trends and patterns, capture complex relationships and find external factors influencing demand. Model creation will enhance
the accuracy of demand, inventory management, and improve the decision-making process of the company. The future outcome of
this project, the company will be able to meet customer demands in a better manner, reduce inventory costs, and maintain a
competitive edge in the market.
Keywords—Demand forecasting, Biscuit manufacturing, Inventory management, Machine learning.
I. Introduction
The biscuit manufacturing industry is a dynamic sector characterized by a diverse range of products and intense market competition.
Companies under this industrial category provide a wide range of products under different brand names and SKUs (Stock Keeping
Units). So, the biscuit manufacturing industry is considered as a competitive environment in the Sri Lanka as well as the world, in
such competitive environment, the accurate demand forecasting is crucial to maintain an efficient inventory management, to
optimize production process, and controlling the costs. Two of the principal production costs involved are depreciation of the
production line and wastage arising from product which is out of specification. [1] Forecasting demand involves predicting future
product demand based on historical data, market trends and patterns, and other factors influencing. Accurate demand helps the
companies to align their production schedules as well as distribution channels with anticipated demand, this reduces the risk of
overstock and helps to enhance customer satisfaction. The company I focus on is a privately held mass production enterprise
specializing in the manufacturing of biscuits, and along with some other categories such as agricultural and cereal products and
milk products. However, 60% of the company’s production, and sales and distribution are rooted in the biscuit manufacturing
sector, which maintains the company’s stability and growth. With a workforce of over 1000 employees, the company caters for
both local and international markets, producing and distributing biscuits under various brand names. The company’s end customers
are sales representatives and outlet bases, who purchase and distribute the products to consumers worldwide. Sales are considered
as the primary driver of this company’s profit and wealth. In this context, accurate demand forecasting has become important, as it
directly influences the sales outcomes and overall business performance. The demand forecasting process requires handling
sensitive data to predict future market needs accurately, ensuring that the company can meet the customer demand efficiently and
sustain in the competitive global market. Currently, the company manufactures over fifty different biscuit products, each associated
with a distinct brand and SKUs. To forecast demand for these products producing, the company relies on manual methods prepared
by sales and administration, and marketing departments. These methods are conducted with the help of tools like excel spreadsheets,
where the data collected from a system stored and entered in a blank space, manipulated, and analyzed to generate forecast. As this
practice is a traditional and manual practice, it will produce challenges that undermine the accuracy and reliability of the forecast
obtained. The limitation of the current forecasting methods highlighted the urgent need for a more robust and automated approach.
The inefficiency associated with manual forecasting will not only affect the accuracy of demand predictions but also lead to an
increment of operational costs and reduced agility in response to market changes. As the industry continues changing consumer
preferences, market competitions and market conditions, the company must adopt advanced forecasting techniques to stay
competitive, strong, and stable and meet the customer demands effectively. To overcome these challenges, this study aims to
develop predictive analytics forecasting models that leverage machine learning techniques to obtain an accurate demand forecast.
Machine learning (ML) is used to teach machines how to handle data more efficiently. Machine Learning relies on different
algorithms to solve data problems. [2] Utilizing advanced data analysis methods, such as time series analysis, machine learning
algorithms, data preprocessing and feature engineering and some clustering techniques, the model will automate the forecasting
process, thereby will help to minimize the human errors and improve the demand predictions.
II. Methodology
The research methodology is designed to provide a comprehensive approach to develop an accurate and dependable forecasting
model for the biscuit manufacturing company. The methodology comes with machine learning algorithms with robust data
collection and evaluation process to ensure that the predictive analytic conduct is effective or not compared with the current manual
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025
www.ijltemas.in Page 398
forecasting methods. In that case, we apply machine learning. With the abundance of datasets available, the demand for machine
learning is on the rise. Many industries apply machine learning to extract relevant data. The purpose of machine learning is to learn
from data. [2].
A. Approach and techniques: The machine learning model will train with historical sales data and other relevant variables, with the
aim of developing the model that can predict accurately the future demand across various products, brands, and SKUs. The key
steps of the study are data processing, model training, validation, and evaluation. Data preprocessing is a method of cleaning and
organizing the data to ensure the quality of data and its consistency; while doing preprocessing, we can have feature engineering to
create meaningful input variables for the model. Once the data is preprocessed and prepared, the machine learning models can be
trained on a portion of data, while having other portions remaining for validation and testing. The model’s performance is calculated
using some metrics to measure the accuracy and reliability of the model. Finally, the models will be compared against the other
models to identify the best approaching model for the company’s requirements.
III. Results
Exploratory data analysis – Time series of net amount.
The graph shows a time series with the net amount for the product over the time from early 2022 to early 2023. The Y-axis shows
the net amount in millions from 0.8 to 3.5 million, and X-axis shows the month timeline. The blue line depicts the significance of
fluctuations in net amount over the year, with peaks and troughs. The most significant variances shown in the months of March to
July 2022, in April and June 2022 where the graph dips reach close to 1 million. After July 2022, the net amount stayed at a higher
place. As for the future forecast the net amount for the month of January 2023 will be around 3 million with minor fluctuations and
illustrates a slight upwards trend in next following months.
Future predicted net amount values.
IV. Discussion
The findings underscore the limitations of manual forecasting methods currently used within the organization. By leveraging
predictive analytics, businesses can reduce human errors, adapt to changing market trends, and respond quickly to demand
fluctuations. The variation in forecast accuracy across products suggests the need for tailored models rather than a one-size-fits-all
approach. Moreover, incorporating external factors like seasonal trends, promotions, and regional preferences could further enhance
forecasting precision. Future improvements should focus on integrating machine learning algorithms and real-time data to
continuously refine demand predictions.
V. Conclusion
This project aims to improve the demand for forecasting accuracy and analyze the performance of the OREO product line and
provide insight into better decision making. The analysis revealed a forecasting gap, with actual sales of prediction exceeding
45.99%, emphasizing the need for forecasting model. The product named OREO blueberry 240g*4 shows a high performance and
strong demand in the market while OREO real choc and nuts 400g*36 shows a low performance and sales. Some products show
reliable forecasts, others had high error rates. Product performance was categorized as high, medium and low based on total sales,
revealing a strong market. The forecasted summary on the predicted sales and demand shows a high demand for OREO coffee
400g*36 and OREO valentine 200g*25, and OREO oat krunch dark choc 400g*4 with high net amounts in millions. The errors
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025
www.ijltemas.in Page 399
metrics such as MAE, MSE, and RMSE showed a variability in forecast reliability across the products. Overall, the project shows
the importance of having advanced predictive analytics and data-driven strategies to improve allocation of resources, maximize
sales and forecast future data.
References
1. S.C. Cook and A.G. Hall, "Biscuit Baking - A Model Approach," 2005.
2. M. Batta, "Machine Learning Algorithms - A Review," 2019.
3. D. Karthika and K. Karthikeyan, "A Recent Review Article on Demand Forecasting," 2021.
4. K. Teplická, "Using of Optimizing Methods in Inventory Management of the Company," 2020.
5. S. Chopra and P. Meindl, Supply Chain Management: Strategy, Planning, and Operation, 6th ed., Pearson, 2016.
6. E.A. Silver, D.F. Pyke, and D.J. Thomas, Inventory and Production Management in Supply Chains, 2017.
7. M. Armstrong, "Factors Influencing Worker Motivation in a Private African University: Lessons for Leadership," 2012.
8. R. Fildes and P. Goodwin, "Good and Bad Judgment in Forecasting: Lessons from Four Companies," 2007.
