Flavor Forecast: Optimizing Potato Chips Production through

Demand Forecasting Using Machine Learning Techniques

Vishnu Priya R., Jaianandakrishnaa K., Shigivahan A.,

Charan Vivek Raj R. and Alagu Veera Siranjivee D.

Department of Computer Science and Engineering, Kalalsalingam Academy of Research and Education, Anand Nagar,

Krishnankoil, Tamil Nadu, India

Keywords: Predictive Analytics, Demand Forecasting, SARIMAX, Random Forest, Inventory Management, Streamlit,

Time Series Analysis, Cross Validation.

Abstract: In this paper, a comprehensive predictive analytics framework for demand forecasting of fast-moving food

products is proposed with emphasis on potato chips across different flavours. The study uses a hybrid

approach of SARIMAX (Seasonal Autoregressive Integrated Moving Average model extended to include

exogenous factors) and Random Forest algorithms to model historical sales data and exogenous factors such

as sports events, festivals, holidays, and promotional activities. The framework's architecture incorporates

several components: seasonal pattern analysis for the time series, event study to analyse demand shocks and

feature construction for improved model fit. These dynamic factors are incorporated to produce detailed

demand forecasts and indicate the percentage change in demand trends. This approach to production planning

is ahead of time to avoid overproduction, minimize waste and achieve proper inventory management

throughout the supply chain. The implementation is a web application developed using Streamlit, with strong

user authentication, data handling, and visualization features. The system architecture consists of data

ingestion, preprocessing, feature selection, model training, and real-time prediction generation modules. The

preprocessing pipeline contains data cleaning algorithms, temporal aggregation and outlier detection that are

applied automatically to the data. The SARIMAX model was found to be more accurate in providing point

forecasts of demand in real time with an accuracy of 91.47%. This is because it can incorporate both the

seasonal components and other variables easily. The framework’s effectiveness was then established through

a rigorous cross-validation procedure and the use of standard performance metrics such as MAE and RMSE.

1 INTRODUCTION

Potato chips industry needs demand forecasting

solutions that can be applied to complex challenges

arising from consumer behaviors, regular patterns, and

adjusting to seasonal and cultural trends. In 2023, the

global potato chips market was worth 35.23 billion

USD and projected to grow to 49.07 billion USD as of

2032, all while facing massive operational challenges

in optimizing production.

Excess production led to unnecessary resource

utilization and heightened operational costs, however

lack of production during peak seasons resulted in

customer displeasure along with brand loyalty and

market share losses.

We propose that due to the rapidly changing

character of these variables, traditional forecasting

approaches do not work, data-based solutions,

concerning parameters related to the event, will have

to be complex.

To address this inadequacy, this paper presents an

advanced predictive analytics framework that

employs SARIMAX models with granular details

about events to predict demand for various potato

chip flavours. The Framework employs state-of-the-

art machine learning techniques with automated

feature generation and robustness metrics like Mean

Absolute Error (MAE) and Root Mean Square Error

(RMSE). So, SARIMAX models ameliorates the

evergreen and transitional time series models to

estimate the typical trends and abrupt variations

accordingly, carrying the technical baseline at

manage. Advanced statistical methods can be

employed to use multiple data sources to have a better

understanding of demand patterns.

676

R., V. P., K., J., A., S., R., C. V. R. and D., A. V. S.

Flavor Forecast: Optimizing Potato Chips Production through Demand Forecasting Using Machine Learning Techniques.

DOI: 10.5220/0013941800004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 5, pages

676-683

ISBN: 978-989-758-777-1

This system functions as a user-friendly web

application providing data privacy and easy

accessibility while facilitating real-time demand

forecasting and production planning. The framework

addresses critical industry issues such as sustainable

sourcing, inventory optimization and demand

volatility.

The system aligns production planning with real

time demand patterns to ensure inventory levels at all

times, thus improving efficiencies and metrics for

customer satisfaction while enhancing sustainability

initiatives in the food and beverage sector. Overall,

this method accounts for both demand (projects)=

3.75% CAGR market up to 2032 and supply to

demand action will be continued by this robust

"demand forecast", which is a progressive supply

chain supply update. Contextualizing the

methodological construct, empirical findings, and

actionable insights with strategic perspectives, the

framework differentiates its contribution to methods

of demand forecasting with a particular emphasis on

the dynamic market environment of the snack food

sector.

2 RELATED WORKS

In recent years, particularly for the food industry,

machine learning and predictive analytics have

increasingly been utilized for demand forecasting. AI-

based demand forecasting, thus, has drastically

influenced food supply chains with improved

accuracy and reduced food waste (Kumar, S and

Singh, M. 2024). MDAs are considered powerful tools

in demand forecasting and have shown more

flexibility than traditional techniques in coping with

fluctuations in the data (Aci, M and Yergök, D.

2024).

Predictive analytics frameworks combine

statistical methods and advanced technologies and

algorithms that have been shown to enhance food

safety risk management by anticipating risks and

intervening before they develop into a serious problem

(Chen, H and Wang, L. 2025). Over the years, food

demand forecasting has evolved from traditional crude

methods to automated approaches to analyses large

data sets with the incorporation of machine learning

algorithms (Wang, H and Li, Y, 2024).

The application of Random Forest algorithms for

food demand prediction has been found to be

successful in the context of both regression and

classification tasks and for various types of data. The

demand variation is attributed to price changes,

promotions, customer preferences, and weather

conditions, making the prediction especially difficult

for items with short shelf life (Martinez, C and

Rodriguez, P, 2024).

The last area of supply chain predictive analytics

has come up with an improved fleet performance

optimization through the analysis of shipping routes,

weather, and traffic patterns (Brown, R, 2024). Of the

advanced demand forecasting tools have been found

to be critical to F&B firms to better manage their

merchandising and improve their supply chain

management above the manual level.

3 PROPOSED METHODOLOGY

Figure 1: Workflow Diagram of the Preliminary Model.

Figure 1 demonstrates the methodology of the study

as well as the steps involved in the process and the

iterative process for the continuous

enhancement as shown below. The methodology of

demand forecasting for potato chips production is

cyclical and consists of eight interconnected phases

that make up a continuous improvement loop.

Flavor Forecast: Optimizing Potato Chips Production through Demand Forecasting Using Machine Learning Techniques

677

Step 1: Demand Forecasting Initiation.

The process begins with setting the forecasting

requirements' scope, determining the forecasting

intervals, and aligning the objectives with the

organizational goals, such as inventory optimization

and waste reduction.

Step 2: Data Acquisition.

This phase includes the collection of a complete

dataset that comprises historical sales data at various

temporal frequencies, knowledge of seasonality and

trend, information on events such as sports events,

festivals, and holidays, and other market-specific

factors that can affect demand.

Step 3: Data Preparation.

The framework uses stringent data preprocessing

methods, which include data cleaning methods

(Anderson, M and Wilson, K , 2024), the handling of

missing data and outliers, normalization and

standardization of the data, and the creation of

temporal features.

Step 4: Model Development.

Two main predictive models are used: the SARIMAX

model that includes seasonal components and

exogenous variables; and the Random Forest

algorithm which is suitable for modeling non-linear

relationships and interactions between features

(Nassibi, N and Fasihuddin, H, 2023).

Step 5: Model Evaluation.

The model is evaluated using multiple metrics such

as MAE and RMSE (Lee, S., & Kim, J, 2024).

Moreover, grid search is used for model parameter

tuning and cross-validation for model robustness and

generalization.

Step 6: Demand Prediction.

It entails forecasting both the short-term and the long-

term demand, with event-specific demand

fluctuations, and confidence intervals for the forecast

accuracy.

Step 7: Production Alignment.

This phase focuses on the synchronization of

production schedules with the forecasted demand,

inventory management and control, and waste

management.

Step 8: Performance Monitoring.

The last phase of the process is to continue with the

improvement by monitoring the accuracy of the

forecast in real time, updating the model with new

data, analyzing the performance metrics, and

incorporating the feedback to enhance the system.

The demand forecasting methodology for potato

chips production is broken down into eight phases of

a cyclical process. The process begins with

forecasting initiation to determine scope and

objectives and continues through the comprehensive

data acquisition of historical sales data and event

related information. The framework applies strict data

preparation techniques followed by the development

of SARIMAX and Random Forest models. Once

models are evaluated with metrics like MAE and

RMSE, short- and long-term demand predictions are

produced. The process ends with production

alignment and the continuous performance

monitoring of the system to create a feedback loop for

the ongoing enhancement and optimization of the

system.

4 RESULTS AND DISCUSSION

The adoption of the predictive analytics framework

for potato chips demand forecasting yielded rich

insights from multiple visualization analyses that

showed how well the model seized demand patterns

and relationships.

Figure 2: Pie Chart Visually Represents the Sales Volume

Distribution.

In Figure 2, the pie chart reveals the market share

distribution across different potato chip flavors. It

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

678

also shows that the dominant segments are Barbecue

with about 26.7% of sales volume, Classic at 17.7%,

and a Spicy Variant at 16.2% IMARC Group (2024).

This distribution pattern can be seen as strong

consumer preferences for traditional and spicy flavors

that will affect production planning and inventory

management decisions.

Figure 3: Event Type by Demand Volume (Boxplot).

In Figure 3, the boxplot of event-type demand

volumes offers valuable information on demand

variability as a function of event type. The highest

median demand is seen in sports matches and there is

a substantial extension of the upper quartile which

suggests substantial demand spikes during sporting

events PredictHQ Research Team (2024). Festival

periods show consistent demand patterns with

moderate variability and the widest interquartile

range, which suggests less predictable demand

patterns, are seen in the holiday periods.

Figure 4: Demand Volume vs Year (Line Chart).

In Figure 4, the longitudinal analysis from 2014 to

2020 can be represented in a line chart where clear

cyclical patterns of demand volume are visible. The

visualization has some peaks and troughs with

moderate to high demand spikes at certain periods.

These temporal patterns were also well captured by

the SARIMAX model that incorporated both seasonal

components and exogenous variables (Arunraj, N.S.,

& Ahrens, D, 2016).

Flavor Forecast: Optimizing Potato Chips Production through Demand Forecasting Using Machine Learning Techniques

679

Figure 5: Correlation Between Sales Volume and Demand Volume.

In Figure 5, the heatmap visualization shows

strong correlations among the key variables in the

demand forecasting model. First, there is a very high

positive correlation (0.957) between sales volume and

demand volume, which supports the basic

assumptions of the model. The heatmap also helps to

understand the relationships between event types and

demand patterns and reveals that sporting events are

the most correlated with demand spikes. This multi-

variable correlation analysis is very useful for feature

selection and model optimization.

Figure 6: Sales Volume vs Year (Bar Chart).

Figure 6 illustrates the sales volume trends over

the years, highlighting fluctuations in demand from

2000 to 2022. Significant peaks are observed around

2005, 2010, and 2015, with another notable surge in

2020. The trend suggests possible external influences,

market shifts, or seasonal variations affecting sales.

Post-2020, there is a declining trend, indicating a

reduction in sales volume, possibly due to changing

economic conditions or shifting

consumer preferences. In Figure 7, the comparison of

forecasting models shows severe limitations in their

predictive capabilities. Three models, XGBoost,

Random Forest, and LightGBM, show the same

pattern of predictions in the line graph, with only a

steep rise on the end. This lack of variation and non-

response to temporal patterns suggests that the

models fail to capture the complex demand patterns

of the potato chip sales. However, the SARIMAX

model, which was used in this study, was better in

incorporating the seasonal components as well as the

external variables and it provided a very high

prediction accuracy of 91.47%. This makes

SARIMAX a more suitable model for demand

forecasting in this case.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

680

Figure 7: Comparison of XGBoost, Random Forest, and LightGBM.

Figure 8: Sample Forecasting.

In Figure 8, the baseline demand forecast for the

next 30 days is relatively flat with some fluctuations

indicating the demand trends in the absence of other

factors. Nevertheless, the adjusted demand forecast

that includes the effects of certain factors shows more

graded variations and more realistic demand patterns.

The inclusion of event-based factors and seasonal

components lead to a better projection of the expected

demand changes and shifts in the subtle peaks and

troughs of the adjusted forecast line.

Flavor Forecast: Optimizing Potato Chips Production through Demand Forecasting Using Machine Learning Techniques

681

4.1 Model Integration and

Performance Metrics

The success of the framework is based on the use of

multiple data sources and feature types, the robust

handling of seasonality through SARIMAX, the

ability to capture non-linear relationships through

Random Forest, and a comprehensive performance

evaluation against multiple metrics.

4.2 Practical Implications

The visualization results translate into several

practical applications:

4.3 Production Planning Optimization

Accurate short-term demand forecasts enable better

production scheduling, which in turn results in

improved inventory management using better

demand predictions. Thus, by optimizing production

quantities, waste can be greatly minimized, and,

therefore, the manufacturing process can be made

more efficient and cost-effective (Nagaraj et al.,

2023).

4.4 Resource Allocation

This ensures that production resources are distributed

properly and that capacity is used more effectively

based on predicted demand. This approach enhances

overall supply chain management and results in a

more effective operation.

4.5 Event-Based Strategy

Taking advantage of flexible production and being

able to react quickly to production changes helps

organizations to respond better to supply chain

events and demand fluctuations that come with

seasonless. So, with this, optimizing inventory levels

during peak demand durations, the corporations will

now no longer best retain the essential stages of

effectiveness but additionally follow requirements in

catering to patron needs, e-journal of technology &

engineering.

The experimental results demonstrate the

effective performance and practicability of the

proposed framework to address the primary

challenges of potato chips demand forecasting, and

the resulting data will enable you to gain productive

insights for production planning and inventory

management. The model demonstrates significant

potential for optimizing production schedules and

reducing waste, which can improve sustainability in

the snack food sector.

5 CONCLUSIONS AND FUTURE

ENHANCEMENT

Using the SARIMAX model, which proved to be

91.47% accurate, the research has implemented and

achieved a complete predictive analytics framework

for potato chips demand forecasting. The framework

demonstrates strong capacity to model the demand

trend of the different flavours, and further exploratory

data analysis indicates that Barbecue flavour is most

preferred followed by Classic and Spicy variants

respectively.

Thus, confirming that this SARIMAX model is

suitable for the data as it incorporates multiple

exogenous variables including event-based and

seasonal variables. A large amount of validation was

performed to select the model parameters and results

validated that SARIMAX was in particular much

more effective with external data included.

The framework has shown strong performance as

indicated by the high correlation coefficient of 0.957

between the actual and predicted demand volumes

while compared to traditional forecasting methods.

The implementation of the model in a Streamlit

web application provided the stakeholders with a

user-friendly interface that could be used to get real-

time demand forecasting and production planning.

The architecture of the system allows for easy

integration, preprocessing and visualization of data

which in turn enhances decision making.

This technological integration has been especially

useful in the context of inventory management and

waste reduction. The systematic approach to demand

forecasting has brought about significant

enhancements in the major operational areas:

Accurate stock level predictions have enhanced

inventory management, while optimized production

schedules have led to reduced waste and data-driven

decision making has enhanced resource utilization.

These changes are aimed at increasing the efficiency

of the operations of the snack food industry which is

in line with the current requirements for supply chain

management and environmental sustainability. The

framework's capacity to integrate complex statistical

analysis with user friendly interfaces makes it suitable

for further development and application in the food

and beverage industry, for products that have similar

demand patterns and shelf life.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

682

Future work could include applying machine

learning algorithms such as LSTM and Prophet to

compare results, incorporating real time market

sentiment analysis and creating automatic production

scheduling systems. Moreover, the ability to include

price elasticity analysis and competitor data could

improve the accuracy of the forecast. The

implementation of blockchain technology for supply

chain transparency and the integration of IoT sensors

for real time inventory tracking could improve the

performance of the system even further.

REFERENCES

Aci, M., & Yergök, D. (2024). "Demand Forecasting for

Food Production Using Machine Learning

Algorithms." Technical Gazette, 31(1), 123-134.

Anderson, M., & Wilson, K. (2024). "Accurate Demand

Planning in F&B Industry." ThroughPut Analytics

Journal, 4(1), 112-126.

Arunraj, N.S., & Ahrens, D. (2016). "Application of

SARIMAX Model to Forecast Daily Sales in Food

Retail Industry." International Journal of Operations

Research and Information Systems, 7(1), 1-21.

Brown, R. (2024). "Predictive Analytics for Reducing Food

Waste." The Rail Media Research, 6(4), 178-192.

Chen, H., & Wang, L. (2025). "Leveraging Predictive

Analytics for Food Safety Risk Management." Journal

of Industrial Production Development, 9(1), 78-92.

Davis, M., & Miller, S. (2024). "Machine Learning-based

Demand Forecasting Against Food Waste." Journal of

Industrial Ecology, 28(1), 67-82.

IMARC Group. (2024). "Global Potato Chips Market

Report 2024-2033." Market Research Report.

Johnson, K., & Williams, P. (2024). "Time Series

Forecasting in Food Demand Supply." IEEE

Transactions on Industrial Applications, 60(4), 456-

470.

Kumar, S., & Singh, M. (2024). "AI-Based Demand

Forecasting in Food Supply Chain Management."

Journal of Supply Chain Analytics, 15(2), 45-62.

Lee, S., & Kim, J. (2024). "Effective Food Demand

Forecasting Using Machine Learning." IEEE

Transactions on Industrial Informatics, 16(8), 345-358.

Martinez, C., & Rodriguez, P. (2024). "Demand

Forecasting Models in Food Industry." University of

Minho Research Repository, 8(2), 145-160.

Nagaraj, P., Raj, R. C. V., & Shigivahan, A. (2023,

December). Data Visualization and Analytics for Price

Elasticity on Commodities Using Machine Learning

Techniques. In 2023 International Conference on Data

Science, Agents & Artificial Intelligence

(ICDSAAI) (pp. 1-5). IEEE.

Nassibi, N., & Fasihuddin, H. (2023). "Demand Forecasting

Models for Food Industry by Utilizing Machine

Learning Approaches." International Journal of

Computer Applications, 14(3).

PredictHQ Research Team. (2024). "Event-Driven Demand

Forecasting in Food Industry." Supply Chain Analytics

Quarterly, 8(2), 112-126.

Thompson, J. (2024). "Predictive Analytics Applications in

Supply Chain Management." DiLytics Research

Papers, 5(2), 67-82.

Wang, H., & Li, Y. (2024). "Machine Learning Approaches

in Food Industry Demand Forecasting." International

Journal of Supply Chain Management, 11(3), 234-248.

Flavor Forecast: Optimizing Potato Chips Production through Demand Forecasting Using Machine Learning Techniques

683