IEEE Consumer Electronics Magazine - March/April 2023 - 115

the given unknown land. Recognizing the right
amount of fertilizer for applying on the land
protects the soil wealth and prevents soil
erosion.
EXPERIMENTAL SETUP AND RESULTS
We analyze the performance of all four prediction
models in this section: gradient boosting
regression trees, random forest, convolution
neural network, long-short term memory, and BiLSTM.
The data extracted by sensors are large in
number. So, the deep learning models were
implemented to match the results with experts'
advice in less computational time. These models
were written in Python and tested on a dataset
of soil nutrients.
The learning models use available nutrients,
crop cultivation, and recommended fertilizers as
input data. This case study examined entire
training, testing, and validation data for model
implementation. These data are fed into each
prediction model and then run with standard
parameter settings.
After that, ten runs are used to obtain the
average value of the metrics determined. The
below graph shows the comparison of all taken
algorithms by varying training-testing ratios (see
Figure 3).
Finally, it is important to notice that the BiLSTM
prediction model performs effectively and
produces better outcomes. The preceding comparison
graph and numeric representation of the
metrics with different model comparisons
expressed the amount of fertilizer forecast close
to the expert's advice. As a result, the purpose
of this comparison research is to assist farmers
in increasing their production without relying on
specialists.
CASE STUDY: IOT SENSOR DATA
INTEGRATED SMART MOBILE
APPLICATION
The lack of information about today's technologies
is a huge disadvantage for farmers.
They rely on agricultural specialists for advice
on how to always boost crop yields on their
property. As a result, an application with modules
that meets their fundamental demand for
farming has been developed to help them
March/April 2023
improve their financial situation by boosting
crop production. All required for the application
is information on their land's soil nutrients.
Farmers must feed the application with the value
of soil nutrients to process module output. The
lack of information about today's technologies is
a huge disadvantage for farmers. They rely on
agricultural specialists for advice on how to
always boost crop yields on their property. As a
result, an application with modules that meets
their fundamental demand for farming has been
developed to help them improve their financial
situation by boosting crop production. All
required for the application is information on
their land's soil nutrients.
Sing-in Module: The users can directly sign in
with the application can proceed with the main
module. When users attempt to register themselves,
the application expects them to provide
their " location on permission " to collect their
exact details.
Home Page: The above screenshot is the
" Home Page " of the Kisaan Sasya Application,
which comprises all five modules with different
understandable icons (see Figure 4).
Intro - Slides: Dynamic banners that display
the new idea of this application make greenish
crops with higher yields.
Fertilizer Totalizer: Apart from the information
on nutrient management detailed in the
knowledge base, the app contains information
on the recommended dose of fertilizer for all the
crop varieties concerning growing states (see
Figure 5).
This module expects farmers to input the
land size in acres or hectares.
Based on the land size and already given
nutrients data, it will suggest a minimum of
three combos of fertilizers as DAP/MOP/
UREA for the land.
Based on the farmer's financial status, they
can opt for any combos of fertilizers.
Based on the farmer's financial status, they
can opt for any combos of fertilizers.
Before coming to the fertilizer module, it is necessary
to do a health check for every farmer's
land using the health check module.
So that, based on the entered nutrient value
and suggested crop, this module recommends
115
IEEE Consumer Electronics Magazine - March/April 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March/April 2023
