A Mobile App for Food Purchase Decision and Waste

Minimizing Using IoT, Social Tools, ML and Chatbots

Robin Faro

, Angelo Fortuna

and Giuseppe Di Dio

Deepsensing SRL, Innovative Startup, Catania, Italy

Department of Electrical and Computing Engineering, Catania University, Italy

CERISVI, Research Center for Innovation, Catania, Italy

Keywords: DSS, Chatbot, Social Media, Agri-Food Retail, Mobile App, Human-System Interactions.

Abstract: Chatbots and conversational systems are increasingly emerging as technologies to support decision- making

systems and to improve human-machine interaction. Our paper aims to demonstrate how social media and

chatbots can improve the decisions of a consumer of food products and reduce food waste, whereas simplified

conversational systems are taken into account to facilitate the interaction between users and application. In

particular the paper presents a mobile app for Food Purchase Decision and Waste Minimizing where social

tools and chatbots play an important role to support the implementation of an electronic pantry to optimize

food purchase and consumption. This smart pantry has a memory of all the foods present in the home pantry.

This allows the app to recommend the use of products that are about to expire, to provide with the help of a

chatbot advice for the purchase of products useful for making recipes that take into account the products

present in the pantry, to highlight gastronomic events to participate in for the type of tastings that are offered.

1 INTRODUCTION

Chatbots and conversational systems are increasingly

recognized for enhancing decision-support systems

(DSSs) (N. Bhaskar, 2023; D. Albuquerque, 2021)

and improving human-machine interaction (A.F.

Fujii, 2023). Recent research emphasizes developing

efficient conversational systems (P.D. Sree, 2023)

and showcasing the role of social media and chatbots

in specific applications (C.H.S. Pokhariya, 2024).

This paper explores how IoT, social media and

chatbots enhance food purchase decisions and reduce

waste. It introduces a mobile app for Food Purchase

Decision and Waste Minimization, integrating an

electronic pantry to optimize food management. The

pantry tracks home food inventory, recommends

using expiring products, suggests recipes, and

highlights nearby gastronomic events.

The chatbot plays a crucial role in providing

actionable recommendations and fostering

sustainable consumption. It suggests recipes that

minimize food waste by utilizing pantry items. By

integrating with the app, the chatbot helps enhance

decision-making related to food purchases and

consumptions.

Other key features include seamless product entry

via barcodes/RFID tags and voice-to-text systems,

with product data stored in relational databases like

MySQL or Postgres. Time-series data is managed

using InfluxDB, enabling trend analysis and

consumption classifications and forecasting through

statistical and Machine Learning (ML) techniques.

The proposed app is a game-changer in the

home food management space. By combining IoT-

enabled automation, conversational DSSs, and

proactive waste reduction strategies, it addresses pain

points that other apps (e.g., Too Good to Go,

Mealime, Paprika, Yummly, and Big Oven) only

partially solve. This positions the app as an original

and competitive solution with high potential to attract

users and make a significant environmental impact.

For enterprises operating in food production

and distribution, the app is a valuable tool because it

aims at significantly reduceing food waste and its

associated carbon footprint, contributing to

sustainability goals. In particular, he app fosters

stronger relationships across the supply chain by

showcasing stakeholders' commitment to

environmental responsibility and social impact. By

enabling real-time data sharing, facilitating surplus

donations, and aligning sustainability goals, it

Faro, R., Fortuna, A. and Di Dio, G.

A Mobile App for Food Purchase Decision and Waste Minimizing Using IoT, Social Tools, ML and Chatbots.

DOI: 10.5220/0013346800003929

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

In Proceedings of the 27th Inter national Conference on Enterprise Information Systems (ICEIS 2025) - Volume 1, pages 1069-1076

ISBN: 978-989-758-749-8; ISSN: 2184-4992

1069

enhances collaboration, trust, and mutual benefits

among producers, distributors, and retailers.

Section 2 outlines the app's basic framework as an

e-commerce system integrated with an electronic

pantry. Section 3 details product data loading with

Bar Code/QR code and RFID tags and voice-text

systems. Section 4 explains how the electronic pantry

supports decisions through consumption statistics and

recommendations. Section 5 focuses on the chatbot's

role in providing recipe suggestions and identifying

relevant events. The conclusion highlights the app’s

benefits and potential future developments.

2 A SMART PANTRY FOR

SMARTER FOOD

CONSUMPTION

The app's initial version integrates a cross-platform

solution (Android/iOS) with web dashboards.

Producers manage offerings, while consumers get

pantry-based recommendations to reduce waste and

suit preferences, with organized inventory views and

category filters.

Producers can manage product details (e.g.,

quantity, price, expiry) and publish food and wine

events. The consumer dashboard allows users to

explore products, events, and manage purchases with

e-commerce features like keyword search, category

navigation, and shopping carts.

Fig.1 highlights the app's food categories,

gatronomic events and smart pantry, designed to help

users manage purchases by tracking expiration dates

and usage. Features include: a) adding products via

QR/barcode scanning, b) viewing all pantry items,

opened or closed, c) tracking expiration and post-

open eding deadlines, d) receiving alerts for near-

expiry items, and e) creating shopping lists with

recommendations.

Fig. 2a shows the smart pantry interface, accessed

via voice commands or product tag scanning,

displaying pantry items linked to detailed product

sheets (e.g., name, description, images, and

attributes). Fig. 2b aids purchase decisions by

indicating product status (opened/closed, quantity,

expiration) with color-coded indicators.

The app structure is therefore a step ahead of

traditional e-commerce apps as it effectively

integrates the process of purchasing with the

storing/consuming food products in order to support

sustainable food consumptions.

Figure 1: Smart pantry interface accessed via voice

commands or tag scanning, showing items linked to

detailed product sheets (e.g., name, description, images).

Figure 2: List of products in the pantry and their status.

However, there are some limitations that reduced

the effectiveness of the initial version of the app only

to the purchase and consumption of local products

sold through the app.

smart pantry

categories

events

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

1070

In fact only for these products it is reasonable to

assume that the producers may accept to use our

QRcode/RFID tag in which they specify for each

batch the expiration date and the period within which

to consume the products once opened.

Moreover, upon subsequent analysis, it turned out

that the consumer generally does not want to take into

account only the typical products in the pantry but all

the products to have a complete picture of the food

consumption.

Therefore, it seemed useful that the pantry, like

similar apps available on the market (M. M. Khan

2020, Studio56 2022) will be able to manage products

purchased from other commercial sites or directly at

the supermarket.

Furthermore, given the importance that

consumers increasingly give to the quantity and

quality of food consumption, it is useful that the

consumers are notified not only when a product is

about to expire but also that they may know the

consumption statistics, possibly including nutrients,

and the pleasant and compatible ways of

consuming the available products, especially those

about to expire, with diets and economic criteria.

For this reason, the following sections show the

generalization of the electronic pantry currently

offered by the app using social tools and the

enhancement of the interface with conversational

systems and chatbots for an optimized use of food

consumption and for a minimization of waste in a non-

imposing but useful and pleasant way.

3 IoT AND SOCIAL TOOLS FOR

SMART PANTRY DATA ENTRY

Despite the growing number of IoT devices and

technological integrations in households, pantry

management often remains a largely manual process,

discouraging monitoring of the products on the

pantry. To address this, the paper proposes a

rethinking of product labelling, so that labels contain

not only the name and price but also a range of

parameters that allow interaction with the product

itself.

For this reason, it is useful to design systems that

facilitate the insertion of data for new products into

the pantry DB, in our case Postgres DB, avoiding or

limiting the intervention of the consumer to enter such

data via the keyboard. Furthermore, these data could

not be limited to the bar code or QRcode, but should

provide the name and the quantity of the purchased

product, the nutritional contents and the expiry date

for an optimal consumption.

As above said, a first method to avoid this

criticality is to provide each product with our tag able

to store all the information mentioned above, which

would be partly stored in the tag and partly obtainable

from a database containing all the food products

managed by the app. But replacing the bar code of

their products with this QRcode/RFID that facilitates

is reasonable only for local producers as shown in

Fig.3 (left).

A possible solution to overcome this problem may

be that the producers themselves or the points of sale

to replicate the bar code inside our tag where they will

insert at least the product expiration date. The product

would thus have both the bar code as long as it is

essential for checkout operations and the

QRcode/RFID tag.

However, also this is difficult to achieve in

practice. Therefore, in the following we assume that

the products at e-commerce sites and local stores are

provided only by bar code, as shown in Fig.3 (right).

Figure 3: Products inserted/retrieved using QRcode (left) or

Bar code /right).

For this reason, the consumer, when inserting the

product into the real pantry, should opened a pantry

loading procedure offered by the app and click on the

bar code of the purchased products. If the app does not

find this product in the app's DB, then it would do a get

to a database containing all the products asking for the

data of the product whose identification code is given

by the bar code.

The retrieved data would then be loaded into the

app's DB and in the next screen the consumer would

only have to enter the number of packages purchased

of the same product and the expiry date.

A Mobile App for Food Purchase Decision and Waste Minimizing Using IoT, Social Tools, ML and Chatbots

1071

3.1 Uploading Products Data Using

Bar Codes and Open Food Facts

The previous method allows adding new products to

the app database, assuming a comprehensive food

product database is available. To leverage existing

resources, we identified the Open Food Facts

database (https://world.open foodfacts.org/), a

collaborative project that includes extensive food

product and nutritional data, available for download

and remote querying. Below, we outline the steps to

implement this method, showcasing its effectiveness

through social systems.

The first step requires the customer to open the

Open Food Facts app and select the products of

interest by entering the product name or product type,

thus obtaining a list from which to identify the one of

specific interest.

Fig. 4 (left) illustrates how a consumer can

resolve doubts about a product name by entering the

product type or producer. For instance, if unsure

about a specific pesto by Campo d'Oro, the consumer

can first select the producer and then choose the

desired product from a list provided by the Open Food

Facts app.

Figure 4: List of products of “Campo d’oro” stored on

Open Food Facts (left) and list of products inserted into the

list home from customer (right)

This list, called home in fig.4 right, can be

updated continuously, and each time it is updated it

must be exported to Google Drive where. as shown in

fig.5, it consists of an https link containing a list of

barcodes relating to the selected products.

Figure 5: Bar codes of the products present on the home list.

The mobile app can read this list and query Open

Food Facts with subsequent queries that will provide

the product data that will be uploaded to the pantry DB

with a quantity of zero. This will allow the customer

to just click on the product barcode at checkout to enter

it and its quantity into the pantry. If the product was

not selected at the initial stage using the above

procedure, at the time of purchase the app will

select it from Open Food Facts and insert its

descriptive data into the pantry DB. In case it is not

present on Open Food Facts the app will opened a

screen to insert the product basic data both into the

app DB and into the Open Food Facts DB.

This latter insertion phase is supported and

encouraged by Open Food Facts itself but, for now,

is not available to customers. It will be managed by

the app's team upon the consumer's request.

Additionally, instead of or alongside Open Food

Facts , we plan to utilize product databases from food

distributors adopting the app.

3.2 Inserting Product Quantity and

Expiry Date

After entering product descriptive data into the app's

database, the next challenge is adding quantity and

expiry date information. To address this, customers

utilize a voice-to-text tool that converts spoken

commands into text, which is stored on Google Drive.

The commands, listing items (see Fig. 6.a), are

processed using the Sound-Type-AI converter (see

Fig. 6.b). The app then links each item's data to its

corresponding product barcode.

Alternatively, customers can record the list after

scanning all pantry products without interruptions, a

mobile app matches the scanned product names with

its database, ensuring accuracy. It then records the

quantity and expiration date. If a product is missing

from the pantry database, it is retrieved from Open

Food Facts as described earlier.

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

1072

Figure 6: List of products and relevant data to be

stored in the pantry. A voice-to-text can facilitate this

insertion.

To support consumption and food waste detailed

analysis, the app logs each product into a time-series

database, InfluxDB (X. Zhu, 2023), to track

consumption history for each item (see Fig. 6),

enabling calculations for consumption and waste.

Figure 7: History of four products: P

1,1

and P

1,2

(type P

2,1

and P

2,2

(type P

2).

Vertical lines show remaining

pantry quantities, T

EXP

and T

EXP

are the expiration times

and T

indicates when customers inquire about

consumption and waste. The consumption percentages are

75% for P1 and 25% for P2.

4 CONVERSATIONAL DSS FOR

SMART FOOD CONSUMPTION

For the optimal use of the purchased products, it may

be useful for the customer to know at least the

following data:

which products are expiring and related

warnings/alerts

what is the last weekly or last monthly consumption

of each product

what is the waste for each product on a weekly

or monthly basis

what is the foreseen weekly or monthly

consumption or discard of each product.

Such four main queries give raise to a

recommender system supporting customer decision,

i.e., the responses to such queries are at the core of a

DSS for an effective food management at home.

By the information obtained by the first query, i.e.,

query a), the consumer is encouraged by the app to

immediately consume the expiring products.

By knowing the data obtained by the second query

i.e., query b), the consumer will be able to check

whether the quantities consumed are in line with what

was planned, for example on the basis of a diet or

economic criteria, while the data related to the third

query i.e., query c), may be useful for the consumer to

carefully evaluate the purchased products that are

discarded in significant quantities.

By the fourth query, i.e., query d), the customer

may know what is the forecast of consumptions and

discards on the basis of the food consumed in the last

period. This may induce the customer to modify

timely her/his food consumption.

4.1 Product Status: Inquiries, Alerts,

Warnings, and Recommendations

The product status and related warnings or alerts

required for inquiries of the first type (a) are

calculated through an app procedure that queries the

database to identify products nearing expiration. The

app includes an intelligent notification mechanism

that informs users when a product is close to or very

near its expiration date, issuing a warning or an alert,

respectively, as follows:

• Closed Products: Alerts notify users when a

product nears or reaches expiration, prompting

disposal and removal from the pantry.

• Opened Products: Notifications indicate the

recommended consumption period after opening

and suggest disposal once this period ends.

Therefore, in order to have effective information

about the pantry status, the consumer must enter not

only the product quantity into the app DB and

InfluxDB, but also the remaining quantity once the

product has been openeded after having used it. The

remaining quantity of the products can be entered into

the app using the voice-to-text converter or more

simply by clicking on one of the options offered by

the app.

In particular, in the first case, the consumer will

say that the remaining quantity is a lot, medium , a

A Mobile App for Food Purchase Decision and Waste Minimizing Using IoT, Social Tools, ML and Chatbots

1073

little, or nothing, while in the second case, the

customer may select one of these options by clicking

on a button offered by the app for communicating the

product remaining quantity.

The app computes every day an expiry table T

in which it reports the products that are close to

expire. Let say t

and t

respectively the time at

which the product i is entered into the pantry and has

to be consumed, the remaining quantity r

of the

product will be put in table T

if the last r

stored on

InfluxDB is different from nothing and the

remaining time meets the following formula:

= t

– t ≤ 3 days,

where t is the time at which the computation is done.

Although the expiring products in the next three

days are

communicated by a warning reporting the

table T

, the

customer may issue a voice requests in

natural language to have a prompt response about a

product status. For example, the customer could ask

•

Is milk expiring

•

Is pesto expiring

where the part of the inquiry in bold is mandatory and

the one in italic may contain the name of any product

in the pantry.

Let us note that a warning system implicitly

recommends the product that should be consumed in

the next three days, but it does not indicate at which

rate the consumption should be done, e.g., by

specifying that the consumption should be little

increased or highly increased. For this reason the app

is able to carry out a fuzzy computation for giving the

following more detailed recommendations:

IF the remaining quantity r

is little/medium/a_lot

THEN the percentage consumption c

in the next

day should be:

= 100 x s

/ Ni

where s

is 0.25, 0.5, 0.75 depending on if remaining

quantity

is little/medium/a_lot, and Ni is the product

expiry time expressed in days.

4.2 Inquiries on Consumption, Waste,

and Forecasting

The second piece of information required by inquiry

of type (b) deals with product consumption. It may be

obtained by querying the app with vocal commands as

follows:

•

How much milk have I consumed in the

last week

•

How much meat have I consumed in the

last month

where the part of the inquiry in bold is mandatory and

the one in italic may contain the name of a product in

the pantry and a time period, i.e., week or month,

indicated by the customers.

The response will be the consumption of product

i in the

last week or month, i.e., C

and C

, given

by summing the term (1 - s

) for all the purchased

packages of products in the

last week or month. This

may be easily obtained thanks to data registered in

InfluxDB (as illustrated in fig.6).

A warning is issued by the app reporting the

consumption that exceeds a prefixed threshold, if:

> Th

ciw

or C

> Th

cim

where Th

ciw

and Th

cim

are the consumption

thresholds of

product i for the same period.

Analogously, the third piece of information required

by inquiry of type (c) may be obtained by querying

the app with vocal commands as follows:

•

Were there any discards for apples in the

last week

•

Were there any discards for meat in the last

month

where the part of the inquiry in bold is mandatory and

the one in italic may contain the name of a product in

the pantry and the period chosen by the customer. The

response is given by

summing s

for all the purchased

packages of products i in the

last week or month for

packaged products.Let us note that, to address

questions related to loose products, a possible

simplified approach is to indicate the purchased

quantity qi, expressed in kilograms, for the most

popular items such as bread, meat, and fish when

these products are added to the pantry. In this way, a

warning will be issued reporting the products that

exceeds a prefixed discard threshold, i.e.:

> Th

diw

or D

> Th

dim

where D

and D

is the discard of product i in the

last week or month given by summing the remaining

quantity s

for each product package, or q

x s

for each

loose product, purchased in the last week or month,

and Th

diw

and Th

dim

are the related

thresholds for

the same period. Type (d) questions focus on

forecasting consumption and waste trends, enabling

timely alerts by analyzing historical patterns and

future purchase projections. The app employs three

method such as Linear Regression, ARIMA, and

LSTM, to predict consumption and waste for

upcoming periods using past data (C. Chakraborty,

2017).

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

1074

For food consumption influenced by seasonality,

external factors, and patterns, LSTM often provides

the most accurate predictions, especially with large,

multivariate datasets that incorporate factors like

weather, price, and seasonality. ARIMA is better

suited for simpler, univariate patterns, while

regression is appropriate for data with linear trends.

Fig.8 presents forecasts for next week's meat

consumption and their accuracy, using ARIMA (Fig.

8a) and LSTM (Fig. 8b). As the literature suggests,

we may observe in such figures that LSTM

outperforms ARIMA, particularly for complex and

non-linear patterns. Understanding these trends aids

lifestyle adjustments to meet dietary or economic

goals, with the recommender system suggesting

strategies like gradual food consumption reductions

to achieve targets, offering tailored advice based on

individual needs.

Figure 8: Forecast of the next week meat consumption by ;

ARIMA, and LSTM methods

5 OPTIMIZING SMART PANTRY

WITH A CHATBOT

With growing complexity in food choices and

awareness of sustainability, consumers need

advanced tools for pantry management.chatbot, based

on the Google Gemini 1.5 Flash model with a custom

prompt, provides personalized and proactive support.

The chatbot oversees pantry inventory, monitors

expiration dates and quantities, and suggests items to

use or buy. It recommends recipes using near-

expiration ingredients to minimize waste and tailors

suggestions to user preferences. Fig. 9 (left) shows

the "recipes" feature generating suggestions from

pantry items, while the chatbot option handles text or

voice requests (Fig. 9, right). Upon receiving a

request, the chatbot processes it (Fig. 10, left) and

delivers responses via text or voice (Fig. 10, right).

Figure 9: Two ways to obtain recipe suggestions from the

Chatbot choosing respectively the option recipe or the one

denoted by chatbot.

Figure 10: Interacting by voice with the Chatbot with

requests expressed in natural language. The Chabot will

respond by both voice and text.

If customers want recipes based on specific pantry

items, they can access the electronic pantry, select the

desired products, and click a button (Fig. 11a, left).

The chatbot then suggests recipes (Fig. 11a, right)

using those products, as shown in Fig. 11b, left. The

customer selects a recipe and receives the details from

the chatbot (Fig. 11b, right).

A Mobile App for Food Purchase Decision and Waste Minimizing Using IoT, Social Tools, ML and Chatbots

1075

Figure 11: How obtaining recipes related to a

specific product.

6 CONCLUSIONS

This paper introduces a mobile app for pantry

management integrated with an e-commerce platform

for food products, leveraging IoT as envisaged in

(C.Mauri, 2021), conversational AI, and chatbots to

enhance user experience. Benefits include

streamlined pantry management, simplified

shopping, reduced food waste, and fresher products.

The recommendation system uses statistical and ML

tools, while the chatbot suggests purchases based on

pantry contents and preferences, integrating data from

price lists and producer catalogs for a reliable

database.The app targets a 20% reduction in

household waste, which, according to UNEP Food

Waste Index Report 2021, could result in: a) 15–20

kg of food saved per person annually in high-waste

regions, b) $300–500 in annual savings per

household, and c) reduced CO2 emissions, supporting

sustainability goals.

Future plans include developing a smart system

with Barcode/QRCode/NFC/RFID readers and scales

for easier product management at home. The

backend, currently hosted on Salesforce, Google

Firebase, and local servers, will be centralized on a

private server. Apps will be distributed via Google

Play and Apple App Store. Additionally, a part of the

backend will be implemented also on edge systems

(e.g., Jetson Nano Orin) for local data storage and ML

computations, along with a telemedicine service (G.

Di Dio, 2024) to assist residents remotely. The overall

system integrates pantry management, health

monitoring, and energy consumption optimization,

offering a holistic solution for modern households.

ACKNOWLEDGEMENTS

This work was carried out as part of the Smart Venues

Project, funded by Action 1.1.5 of the P.O. FESR

Sicily 2014/2020.

REFERENCES

N. Bhaskar, et alii, "An Effective Prediction and

Recommendation System for Smart Farming using

Machine Learning Techniques," 2023 International

Conference on Integration of Computational Intelligent

System (ICICIS), Pune, India, 2023,

D. Albuquerque, G. Bhavani, B. Gaur and S. Kuttipravan,

"A Smart Mobile Application to Boost Grocery

Shoppers Experiential Marketing," 2021 Int-

Conference on Computational Intelligence and

Knowledge Economy (ICCIKE), Dubai.

A. Fujii and J. Kristiina, "Opened Source System

Integration Towards Natural Interaction with Robots,"

2022 17th ACM/IEEE International Conference on

Human-Robot Interaction (HRI), Sapporo, Japan, 2022,

P. D. Sree, M. R. Kokkiligadda, J. Teja and Y. Sandeep,

"Product Negotiation in E-Commerce Website using

Chatbot," 2023 7th International Conference on

Computing Methodologies and Communication

(ICCMC), Erode, India, 2023

C. Bhuvaneswari, et alii, "Implementing AI-Powered

Chatbots in Agriculture for Optimization and

Efficiency," 2024 2nd International Conference on

Intelligent Data Communication Technologies and

Internet of Things (IDCIoT), Bengaluru, India, 2024

M. M. Khan, "Development of An e-commerce Sales

Chatbot," 2020 IEEE 17th International Conference on

Smart Communities: Improving Quality of Life Using

ICT, IoT and AI (HONET), Charlotte,NC, USA, 2020

Studio56, e-Grocy, https://studio56.it/articolo/software-

web-application-erp/2021-09-05/grocy-come-

informatiz zare-la-propria-dispensa, 2022

Zhu, Xiaoxiang et al. “Time Series Database Optimization

Based on InfluxDB.” 2023 International Conference

on Power, Electrical Engineering, Electronics and

Control (PEEEC) (2023)

Chakraborty, C., & Joseph, A. (2017). "Machine learning at

the forefront of time series forecasting." SN Applied

Sciences, Springer.

C. Mauri, P .Pasini, Smart grocery retail: the impact of new

technologies, Egea editor, 2021

G. Di Dio, R. Faro, R. Di Dio, A. Fortuna, “Smart

Telemedicine Services for Students on University

Campus and in Jogging

Areas,” IEEE Int. Conf. on

Internet of Things and Intelligence Systems

(IoTaIS), Bali, Indonesia, 2024

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

1076