2023 3rd International Conference on Smart Generation Computing, Communication and Networking (SMART GENCON)

Karnataka, India. Dec 29-31, 2023

Demand Prediction of Consumer Intention to Buy

2023 3rd International Conference on Smart Generation Computing, Communication and Networking (SMART GENCON) | 979-8-3503-1912-5/23/$31.00 ©2023 IEEE | DOI: 10.1109/SMARTGENCON60755.2023.10442766

Edible Items using Machine Learning Techniques

Hitesh Goyal Ashna Sukhija Kavish Bhatia
Chitkara University Institute of Chitkara University Institute of Chitkara University Institute of
Engineering & Technology, Engineering & Technology, Engineering & Technology,
Chitkara University, Chitkara University, Chitkara University,
Punjab, India Punjab, India Punjab, India
[email protected] [email protected] [email protected]
Kalpna Guleria* Shagun Sharma
Chitkara University Institute of Chitkara University Institute of
Engineering & Technology, Engineering & Technology,
Chitkara University, Chitkara University,
Punjab, India Punjab, India
[email protected]* [email protected]
Abstract— Agriculture is a process of simplifying the food then from the wholesale market it is sold to local shops and
chain in nature and redistributing the resources for animals vendors (retailers) which is then finally purchased by
and plant production. It is the process of nurturing livestock consumers in the market. Over the past four decades, India’s
and growing crops. It consists of preparing animal and plant agricultural productivity has increased rapidly [2][3].
items for human consumption and distributing them to According to the Food and Agriculture Organization (FAO),
marketplaces. A maximum number of textiles and food items India comes in second place as the largest producer of fruits
in the world are produced by agriculture. In India, the and vegetables in the world. However, the availability of
primary source of income is agriculture. Various companies food remains a major concern due to the issues of poor
buy crops from the farmers and directly sell them to the
packaging, absence of sheltered storage, inefficient traders
consumers. As per a study by “The India Express”, there are
90-150 million farmers who rely on agriculture to run the
and many more [4]. The fragmented nature of the food
expenses of their daily lives. Each day there is a huge amount industry leads to inefficiency and losses in the supply chain.
of vegetable waste as their demand doesn’t get predicted due to To overcome these challenges, there are various agri-tech
which a volume of vegetables gets rotten. So to deal with this startups and companies which aim at connecting farmers
issue, there is a need for the development of an automated directly to the consumer through online channels and
model which can accurately predict the demand for vegetables providing businesses with fresh produce in the most efficient
for the next day based on the previous data. In this work, a manner. Fresh fruits and vegetables are directly collected
framework has been developed for predicting the demand for from the farmers [3]. Then, in the warehouse quality is
okra and tomato for the next day, so that the farmers won’t checked and the best ones are packed and transported to the
pluck too many vegetables and also keep track of daily sales. consumers in a limited period. The company buys the
We have used four machine learning models namely, linear required amount of fruits and vegetables as per the demands
regression, decision tree, random forest, and logistic regression of the consumers. Keeping all these factors in mind created a
for identifying the demand for vegetables. In this experiment, Machine Learning(ML) model to predict the demand for
two parameters %error and accuracy have been taken which fruits and vegetables using historical data [4].
shows that the decision tree model performs best with the
highest accuracy of 99.62% and logistic regression has shown ML is a branch of Artificial Intelligence (AI) which
the least accuracy at 85.62%. This work can help various enables systems to make successful predictions using past
edible-based eCommerce companies for predicting the demand experiences [4][5][6]. The system is given the ability to think
for the items for the next day based on the previous data. like humans, with the help of various algorithms and
techniques, on its own [7][8]. In today’s modern time, only
Keywords—Agriculture, machine learning, linear regression, agriculture is a field which lacks technological
decision tree, random forest, logistic regression. advancements. ML can play a big role in making agriculture
more efficient and effective and help farmers in getting better
I. INTRODUCTION
MSP for their crops. ML algorithms can easily and
Agriculture is a primary source of livelihood for the accurately predict the weather and soil conditions, demand
Indian economy [1]. Half of the population in India is for the crop and correct price by comparing it with historical
employed in the agriculture sector which has a great data. When we train an algorithm, based on the historical
contribution to the country's GDP and almost 70% of the dataset and apply new data to it, the output which comes is
rural population depends on agriculture for living their daily called prediction. Demand prediction is a technique that
life. Agriculture is the industry that farmers are a part of and gives the output a requirement of a particular product as per
this industry produces a variety of food products for human the demand of the consumer. And in agriculture, it gives the
and animal consumption [2]. Farmers’ main goal is to feed output of a particular crop required as per the demand of the
the population and make a living expense by producing good consumers. It helps the farmers to select and grow the
crops. They work hard for the whole year and produce good appropriate crop to satisfy the demand. This reduces the gap
quality food for a satisfactory life. This food is purchased or mismatch between the demand and supply of the crop and
from the farmers at a reasonable price and stored in also reduces the wastage of crops.
warehouses or cold storage until it gets a good market price.
From the warehouses, food is transported to the wholesaler

979-8-3503-1912-5/23/$31.00 ©2023 IEEE 1

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SAO PAULO. Downloaded on March 18,2024 at 23:15:16 UTC from IEEE Xplore. Restrictions apply.
 This paper is structured as follows: Section II introduces Authors in [11] have done a semantic analysis of the
various state-of-the-art models for the demand prediction of opinions of experts on crop productivity through ML. The
edible items. Section III summarizes the methodology used study has shown a descriptive analysis to increase the crop
for predicting the demand for okra and tomato. Lastly, the yield and provided the requirement of awareness in the
prediction results of the proposed model and conclusion have current agriculture system using ML. The ML algorithms
been mentioned in sections IV and V, respectively. used are KNN and Naive Bayes. It concluded that Naive
Bayes performs better on a text dataset of agriculture
II. RELATED WORK experts' opinions. An ML model has been developed in [1]
An ML model has been developed in [9] for the for the price prediction of crops using supervised ML
prediction of nitrogen in rice crops. The research contained algorithms. The developed model predicted the crop price
three different datasets for the prediction namely, multi- and forecasted how much the price will be there in the
temporal, multi-source, and multi-scale and applied upcoming 12 months. The model has been used to predict
XGBoost, and various other ML models for the prediction the crop price using decision tree regression. An analysis of
which resulted in the XGBoost model performing best neural networks and regression models has been performed
among other algorithms with the highest accuracy. An ML in [12] to Forecast the demand for food. The research done
model has been presented in [10] to work on the problem of is implementing multiple linear regression and an ANN
farmers like price fluctuations of farmer tools in the market model for predicting the food crops demand which is mostly
and supply-demand of the tools required. The research done used in daily life. Table. I show various ML approaches
used various ML models- decision trees, deep neural used in various different sectors of prediction such as rice,
networks and others, for price prediction and supply and crop, and farmer tools price prediction.
demand for farmer tools. It was concluded that the most
accurate and effective ML model for sharing and renting the III. MATERIALS AND METHOD
farmer's equipment is the decision tree. A review of the price This section of the paper provides an overview of the
prediction of crops has been done in [4]. The research done dataset along with the methodology applied to that for
contains a prediction of attributes like soil condition, weather achieving efficient prediction results.
conditions and demand of the crop in the marketplace to
A. Dataset
predict the price using past scenarios. The research has used
many ML models for different purposes- Linear Regression, The dataset is collected from a new startup company
Decision Tree and Random Forest. Various ML algorithms which directly deals with the demand prediction problem of
have been identified in [2] for the identification of crop yield. fruits and vegetable packets. The dataset contains a total
This survey has identified that artificial neural networks number of 330 data points including four attributes: AVG.
(ANN) and SVM models perform best for the prediction of PRICE, PROD_NAME, PACK_SOLD,
crop yield. A survey on emerging trends in ML has been ORDER_OF_PRODUCT. In the proposed model, the AVG.
done in [3] to predict crop yield and its influence factors. The PRICE, PROD_NAME, and PACK_SOLD have been taken
motive of the research is to explore various ML techniques as the inputs whereas the model predicts the output in the
used for predicting crop yield and solving the food problem form of ORDER_OF_PRODUCT.
in the world. The various techniques surveyed are ANN,
LSTM, RNN, Fuzzy Technique and various other ML B. Methodology
models. The best results were obtained through the Neural This subsection represents the proposed methodology
Network technique, ANN and ANFIS. used for the prediction of the demand for okra and tomato.
TABLE I. COMPARISON OF VARIOUS EXISTING WORKS IN TERMS OF In this work, the preprocessing step has been performed
ALGORITHM AND DOMAIN OF THE RESEARCH WORK on the raw data to make it efficient and useful. In this phase
of the proposed model, the merging and removing of
Ref. Technique Domain attributes have been done manually according to the need of
the output and a single .csv file has been formed. In the
[1] Decision tree Crop price prediction second step modification of the data type of an attribute,
PROD_NAME is done by assigning it a numerical value
[2] ANN and SVM Crop yield prediction
instead of categories. The third phase includes the
[3] RNN, LSTM, ANN and fuzzy Crop yield prediction
replacement of null or empty values in the dataset by the
technique mean/median values of the attributes and then randomizing
data to evacuate the impact of ordered data gathered.
[4] Linear regression, random Crop price prediction Perform further analysis such as visualizing data using
forest and decision tree histograms to understand the data and trends more glaringly.
The train and test ratio of the data has been taken as 80:20.
[9] XGBoost Rice crop prediction After cleaning data into two parts, the first part is utilized to
build the model and the latter is to probe the model, check
[10] Decision tree and Deep neural Farmer tools price the accuracy of the model or examine the error in predicted
networks prediction
values. For further evaluation, the CROSS VALIDATION
[11] KNN and Naive bayes Crop yield prediction technique will be used. In cross-validation, the k-fold cross-
validation has been used where the value of k has been taken
[12] ANN Food demand prediction as 9 as shown in Fig. 1.
In this technique, the train set is further divided into 10
sets. Out of 10 sets, 1 part is utilized to examine the model
and the left 9 are utilized to compose it. Repeatedly, the

2
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SAO PAULO. Downloaded on March 18,2024 at 23:15:16 UTC from IEEE Xplore. Restrictions apply.
 cross-validation technique has been done on alternative sets. output variables. It exhibits the linear relationship, which
Finally, the %error of the values has been identified for signifies how the values of dependent variables are altered
predicting the performance of the models. There are various according to the values of independent variables.
ML models which have been used to identify the prediction
of the demand for okra and tomato. These algorithms are Decision tree regressor: The decision tree regressor
used to train the model to forecast the precise/best algorithm is a supervised learning algorithm whose primary
prediction outcome. The proposed methodology has been motive is to forecast the aimed variables by framing a
shown in Fig. 2. training model[15]. It is a technique based on trees in which
each path originating from the root is represented by a data-
isolated sequence until a boolean value is achieved at the leaf
node[15][16]. The decision tree algorithm is also optimized
if the data set carries the minimum number of nodes and is
properly classified.
Random forest regressor: It regressor is a well-admired
algorithm that pertains to the supervised learning technique
[17]. It can help solve both classification and regression
problems in ML. It mainly revolves around the concept of
ensemble learning, which is used to solve complex problems
by combining multiple classifiers for improving the
performance of the model. In short, a random forest is a
classifier in which numerous decision trees are carried out
on various subparts of a given dataset and takes the mean to
Fig. 1. Cross-validation technique improve the predictive accuracy of the model. Simply, a
random forest makes the prediction from each tree and based
The algorithms used in the proposed work are as follows: on the maximum votes prediction, it will predict the final
output [18][19].
Linear regression: Linear regression is a statistical way
used for predictive analysis [13][14]. It could be explained as
a single-layer perceptron neural layout involving input and

Fig. 2. Proposed methodology

3
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SAO PAULO. Downloaded on March 18,2024 at 23:15:16 UTC from IEEE Xplore. Restrictions apply.
 Logistic regression: It is also associated with supervised prediction and actual attribute shows that the execution of the
learning techniques which are used to predict categorical model is good and the predicted values are so close to the
dependent variables from the given set of independent original values.
variables [20][21]. It is used in various predictive models
such as healthcare, stock movement, crop yield and price Fig. 5 depicts the relationship between the actual value
prediction [22][23][24]. The output of this model must be and predicted values of the demands of okra and tomato. The
categorical or distinct values which are in the boolean form graph shows that the random forest has performed quite good
either 0 or 1, true or false, Yes or No. It gives the values in but some predicted values are too large which is directly
probability form between 0 and 1 instead of showing the affecting the accuracy of the model.
values as 0 and 1. This type of model is similar to linear
regression which is best known for solving and predicting
regression-based challenges while logistic regression works
for solving the problems of classification.
IV. IMPLEMENTATION RESULTS
This section identifies the prediction performance of
various ML algorithms used in this paper which are shown
with the help of illustrations in the graphs below.
In the evaluation phase, various algorithms have been used
to forecast the demand for fruits and vegetables and gained
different results from each algorithm. The proposed model
is based on a regression problem hence there are various Fig. 5. Predicted vs original values using random forest
regression-based parameters which have been utilized to
identify the performance results of the proposed work which Fig. 6 shows the prediction performance of the logistic
are: accuracy and %error. regression model in the case of predicting the demand for
okra and tomato. The graph identifies that the logistic
regression model has a very huge difference between the
actual demands and the predicted demand which results that
this model is not suitable for prediction with such kind of
dataset.

Fig. 3. Predicted vs original values using linear regression

Fig. 3 shows the relationship between the original values

and predicted outputs using the linear regression algorithm.
This graph shows that the model has some variations
between the prediction results and the actual values.
Fig. 6. Predicted vs original values using logistic regression

Fig. 7 shows the comparative analysis of various ML

algorithms in the form of accuracy. The comparison results
show that the decision tree algorithm performs best with the
highest accuracy of 99.62% among all the algorithms for
predicting the demand for okra and tomato. The random
forest model also performs better after the decision tree with
an accuracy rate of 98.61%. Furthermore, the accuracy of the
linear and logistic regression have been identified as 95.90%
and 85.62%, respectively.
Fig. 8 shows the comparative analysis of various ML
models based on %error. This result identifies that the
decision shows the least %error as 0.37 whereas random
Fig. 4. Predicted vs original values using decision tree forest, linear regression and logistic regression perform after
that with the %error of 1.38, 4.09 and 14.377, respectively.
Fig. 4 describes the results of the decision tree algorithm
which performs quite well on the dataset and predicts the
precise values. The comparison between the values of

4
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SAO PAULO. Downloaded on March 18,2024 at 23:15:16 UTC from IEEE Xplore. Restrictions apply.
 several advancements in the realm of agriculture have been
put into practice. With the advent of AI and ML, various
issues which are faced by farmers have been resolved.
Similarly, there is another issue causing a big loss to the
farmers which is the demand for crops and vegetables
required the next day. Various companies buy crops along
with vegetables from farmers and sell them directly to
consumers. But the problem with this is unpredicted demand
due to a huge amount of rotten vegetable and fruit waste each
day. These companies buy only limited stock. However, it
causes problems for the farmers as the remaining vegetable
and fruit get rotten and wasteful. The objective of this paper
is to describe modern ML techniques in the commercial
agricultural sector. To reduce the wastage of fruits and
Fig. 7. Accuracy comparison of various machine learning algorithms
vegetables by predicting the optimum demand of fruits and
vegetables required. In this paper, we are discovering the
solution to a real-time problem in which we predicted the
demand for a particular product, utilizing historical data. ML
plays a vital role in the prediction of the demand for
vegetables and fruit. The visualising techniques of ML are
also used for a better understanding of data, such as graphs
and histograms that provide ease in determining the
performance of the algorithm. Four different ML approaches
namely random forest, decision tree, logistic regression and
linear regression have been used for the prediction based on
the previous data. In this paper, the decision tree has been
identified as the best method with the highest accuracy value
of 99.62% for predicting the demand for crops in the
upcoming day. While the performance of random forest,
Fig. 8. %error comparison of various machine learning algorithms linear regression and logistic regression have been identified
with the accuracy value of 98.61%, 95.90% and 85.62%,
The performance results of logistic regression, linear respectively. This article can help various eCommerce
regression, random forest and decision tree have been companies for predicting the vegetable and fruit demand
tabulated in Table II. as %error and accuracy. each day. Furthermore, this work is also helpful for academia
to understand the applications of various algorithms used in
TABLE II. PERFORMANCE COMPARISON OF VARIOUS TECHNIQUES
PROPOSED IN THE WORK WITH EXISTING WORK the paper.
REFERENCES
Technique Accuracy %error
[1] R. Dhanapal, A. AjanRaj, S. Balavinayagapragathish, and J. Balaji,
Linear Regression 95.9% 4.09% “Crop price prediction using supervised machine learning
algorithms,” J. Phys. Conf. Ser., vol. 1916, no. 1, p. 012042, 2021.
Decision Tree 99.62% 0.37% [2] V. Nathgosavi, “A survey on crop yield prediction using machine
learning,” Turkish Journal of Computer and Mathematics Education
(TURCOMAT), vol. 12, no. 13, pp. 2343–2347, 2021.
Random Forest 98.61% 1.38%
[3] N. Bali and A. Singla, “Emerging trends in machine learning to
predict crop yield and study its influential factors: A survey,” Arch.
Logistic Regression 85.62% 14.37%
Comput. Methods Eng., vol. 29, no. 1, pp. 95–112, 2022.
Naive Bayes[11] 87% 13% [4] M. Rakhra et al., “Crop price prediction using random forest and
decision tree regression:-A review,” Mater. Today, 2021.
KNN[11] 85% 15% [5] S. Sharma and K. Guleria, “Deep learning models for image
classification: Comparison and applications,” in 2022 2nd
International Conference on Advance Computing and Innovative
Technologies in Engineering (ICACITE), 2022.
Table II. presents the performance comparison of the [6] R. Sharma and V. Kukreja, “Mustard Downy Mildew Disease
proposed work and existing work. This comparison shows Severity Detection using Deep Learning Model,” in 2021
that the proposed decision tree model as an outperforming International Conference on Decision Aid Sciences and Application
(DASA), pp. 466-470, 2021.
model having the highest accuracy rate of 99.62% while
other proposed models namely linear regression and random [7] S. Sharma and K. Guleria, “A systematic literature review on deep
learning approaches for pneumonia detection using chest X-ray
forest also outperform the existing models. Furthermore, in images,” Multimed. Tools Appl., 2023.
comparison with logistic regression, it performs quite better [8] R. Sharma and V. Kukreja, “Amalgamated convolutional long term
than KNN[11] but underperformed when compared to Naive network (CLTN) model for Lemon Citrus Canker Disease Multi-
Bayes[11]. classification,” in 2022 International Conference on Decision Aid
Sciences and Applications (DASA), 2022.
V. CONCLUSION [9] M. Iatrou et al., “Topdressing nitrogen demand prediction in rice crop
using machine learning systems,” Agriculture, vol. 11, no. 4, p. 312,
The agriculture sector is extremely important to the 2021.
country's economy in the nation. In the past few years,

5
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SAO PAULO. Downloaded on March 18,2024 at 23:15:16 UTC from IEEE Xplore. Restrictions apply.
 [10] M. Rakhra, A. Bhargava, D. Bhargava, R. Singh, A. Bhanot, and A.
W. Rahmani, “Implementing machine learning for supply-demand
shifts and price impacts in farmer market for tool and equipment
sharing,” Journal of Food Quality, 2022.
[11] M. Rehman et al., “Semantics analysis of agricultural experts’
opinions for crop productivity through machine learning,” Appl.
Artif. Intell., vol. 36, no. 1, pp. 1–16, 2022.
[12] B. P. Bv and M. Dakshayini, “Computational Performance Analysis
of Neural Network and Regression Models in Forecasting the Societal
Demand for Agricultural Food Harvests,” in Research Anthology on
Artificial Neural Network Applications, IGI Global, 2022, pp. 1287–
1300.
[13] W. Wei and X. Yang, “Comparison of diagnosis accuracy between a
backpropagation artificial neural network model and linear regression
in digestive disease patients: An empirical research,” Comput. Math.
Methods Med., vol. 2021, p. 6662779, 2021.
[14] T. K. Saha, S. Pal, and R. Sarkar, “Prediction of wetland area and
depth using linear regression model and artificial neural network
based cellular automata,” Ecol. Inform., vol. 62, no. 101272, p.
101272, 2021.
[15] B. Charbuty and A. Abdulazeez, “Classification based on decision
tree algorithm for machine learning,” Journal of Applied Science and
Technology Trends, vol. 2, no. 01, pp. 20–28, 2021.
[16] C. S. Lee and P. Y. S. Cheang, “Predictive analysis in business
analytics: Application of decision tree in business decision making,”
Adv. Decis. Sci., vol. 26, no. 1, pp. 1–29, 2021.
[17] M. A. Khan et al., “Application of random forest for modelling of
surface water salinity,” Ain Shams Engineering Journal, vol. 13, no.
4, 2022.
[18] V. K. Gupta, A. Gupta, D. Kumar, and A. Sardana, “Prediction of
COVID-19 confirmed, death, and cured cases in India using random
forest model,” Big Data Min. Anal., vol. 4, no. 2, pp. 116–123, 2021.
[19] K. Guleria, S. Sharma, S. Kumar, and S. Tiwari, “Early prediction of
hypothyroidism and multiclass classification using predictive
machine learning and deep learning,” Measurement: Sensors, vol. 24,
no. 100482, p. 100482, 2022.
[20] P. Schober and T. R. Vetter, “Logistic regression in medical
research,” Anesth. Analg., vol. 132, no. 2, pp. 365–366, 2021.
[21] X. Song, X. Liu, F. Liu, and C. Wang, “Comparison of machine
learning and logistic regression models in predicting acute kidney
injury: A systematic review and meta-analysis,” Int. J. Med. Inform.,
vol. 151, no. 104484, p. 104484, 2021.
[22] P. K. Sarangi, K. Guleria, D. Prasad, and D. K. Verma, “Stock
movement prediction using neuro genetic hybrid approach and impact
on growth trend due to COVID-19,” Int. j. netw. virtual organ., vol.
25, no. 3/4, p. 333, 2021.
[23] S. Sharma, K. Guleria, S. Tiwari, and S. Kumar, “A deep learning
based convolutional neural network model with VGG16 feature
extractor for the detection of Alzheimer Disease using MRI scans,”
Measurement: Sensors, vol. 24, no. 100506, p. 100506, 2022.
[24] S. Srivastav, K. Guleria and S. Sharma, "Tea Leaf Disease Detection
Using Deep Learning-based Convolutional Neural Networks," 2023
IEEE World Conference on Applied Intelligence and Computing
(AIC), Sonbhadra, India, 2023, pp. 569-574, doi:
10.1109/AIC57670.2023.10263835.

6
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SAO PAULO. Downloaded on March 18,2024 at 23:15:16 UTC from IEEE Xplore. Restrictions apply.