Crop yield prediction using deep learning.pdf

International Journal of Advanced Science and Technology
Vol. 28, No. 17, (2019), pp. 411-419
411
ISSN: 2005-4238 IJAST
Copyright ⓒ 2019 SERSC
Multi- Model Ensemble with Deep Neural Network Based Crop
Yield Prediction
M.Saranya1
,
1
Research Scholar,
Dept. of Computer Science,
Erode Arts & Science College, Erode-9.
Dr.S.Sathappan2
2
Associate Professor,
Department of Computer Science,
Erode Arts & Science College, Erode-9.
Abstract
Reliable prediction of crop yield is important in the creation of successful regional and
global agricultural and food policies. When production has been predicted, farm inputs like
fertilizers will vary according to plant and soil requirements. Different techniques have been
proposed for accurate prediction of crop yield. Deep Neural Network (DNN) was introduced
to understand the environmental factor i.e., weather data for accurate yield prediction. In
order to enhance the accuracy of yield prediction, Multi-Model Ensemble with DNN (MME-
DNN) is introduced where climate, weather and soil data are considered for yield prediction.
A statistical model is used to find the variation of climate, weather and soil parameters from
year-to-year. These variations are used for climate, weather and soil predictions and these
predictions play an important role in yield prediction. The predicted climate, weather and
soil parameters are given as input to DNN for yield prediction. The yield prediction accuracy
is improved by considering various environmental data.
Keywords— Yield Prediction, Deep Neural Network, Statistical model, Multi-Model
Ensemble.
1. Introduction
The demand of agriculture gets increased because of increasing population in the world
[1]. Agricultural monitoring is faced with specific issues such as seasonal trends related to
plant phenology and crop production depends on climate, environment and soil parameters.
The majority of farmers do not achieve the anticipated yield due to these issues. Farmers want
guidance in good time to forecast upcoming crop productivity, and analysis must be carried
out to assist farmers to maximize production. The yield prediction [2, 3] is one of the main
problems in agriculture. Each farmer would like to know the crop yield. In the past, the yield
prediction was achieved by evaluating the prior experience of the farmer with a similar crop
manually. Nevertheless, the amount of farm data is high and the data evaluation manually is
extremely complex.
Various data mining techniques [4] have been proposed for crop yield prediction to reduce
time complexity and reduce the user involvement in crop yield prediction. For maize yield
prediction, a crop yield prediction method has been implemented using Deep Neural Network
(DNN) [5]. It predicted the crop yield by using different layers. By increasing the number of

Vol. 28, No. 17, (2019), pp. 411-419
412
layers, more features were extracted and it aided to obtain high crop yield prediction
accuracy.
In this paper, Multi-Model Ensemble with DNN (MME-DNN) is developed to obtain
accurate yield prediction by modeling climatic, weather and soil data parameters. In MME-
DNN, year-to-year variation in the climatic, weather and soil data parameters are forecasted
by using statistical model in DNN. The variance of production is estimated using statistical
yield models on seasonal climate, soil and weather conditions. Thus, the crop yield prediction
accuracy is improved by using statistical model in DNN with various parameters.
2. Literature Survey
A crop yield prediction approach based on Adaptive Neuro Fuzzy Inference System (ANFIS)
model [6] was proposed to predict the wheat yield. Initially, biomass, rain, radiation, water
and extractable soil data were collected and it was pre-processed. Then, Neuro Fuzzy,
multiple linear regression and fuzzy logic techniques were processed for wheat yield
prediction. However, ANFIS based crop yield prediction has low Root Mean Square Error
(RMSE) value.
A Bayesian Model Averaging (BMA) [7] was presented to provide a more reliable maize
yield prediction. In BMA, different models such as nitrogen and photosynthesis related
DeNitrification and DeComposition (DNDC) and WOrld FOod STudy (WOFOST) model
and the water oriented AquaCrop model were used to separately produce original country-
level maize yield predictions. After that, a combined prediction was obtained by using a
linear grouping of three ensemble members. The combined prediction model was more
accurate and it was precise predictions of maize yield. The BMA model was remunerated the
uncertainty of individual model effectively. Nonetheless, findings and model data
uncertainties are not fully investigated.
Weighted-Self Organizing Map (W-SOM) [8] was introduced for yield prediction in Mysore
region. W-SOM was an integration of SOM and Learning Vector Quantization (LVQ). The
SOM was employed to reduce the computational cost and reproducible outcome. A fine-tuned
weight factor was included with SOM to enhance the efficiency of weather and crop
prediction. In W-SOM, the accuracy of yield prediction was improved by reducing Within
Class Error (WCE) and range of individual cluster weights. The prediction rate of W-SOM
could be increased by including parallel layer regression with deep belief network strategy for
crop yield prediction.
A crop yield prediction framework [9] was proposed by using Rough Set theory (RS). In this
framework, classification rules from 640 sets of agriculture data was generated by using RS
for crop monitoring. The collected data were pre-processed and then information table and
decision table were generated. Then, reduct was determined by applying a data reduction
method. The reduct holds the minimal set of attributes related to the class label. Finally, the
rules were generated from the reduct by applying LEM2 algorithm. However, this approach
was not suitable for dataset with more number of attributes.
A smart faming system [10] was proposed to predict crop yield per acer. This system
maximized the crop yield by suggesting optimal climatic factors. Initially, yield and weather
data were collected. Then, the most appropriate moisture and temperature data for crop yield
were chosen by applying support vector regression, random forest and multivariate random
regression models. However, this system requires past crop yield data to predict crop yield.
A model based on Fuzzy C Means (FCM) clustering and neural network [11] was proposed
for wheat yield foresee by detecting various parameters like solar radiation, biomass,
temperature and rainfall. FCM allowed one piece of collected data to belong to two or more
clusters that was based on minimizing objective function. FCM predicted wheat yield in

Vol. 28, No. 17, (2019), pp. 411-419
413
terms of degree of membership. It assigning a data point in such a way that it got the close
relationship of similarity between data as much as possible. The clustered data was processed
by neural network which predict the crop yield at a particular region. However, the proper
selection of membership function is more difficult in FCM.
A Convolutional Neural Network (CNN) [12] was proposed for bitter melon yield prediction.
Initially, the descriptions of bitter melon leaf were analyzed using CNN. The description of
bitter melon was given as input to CNN which processed the descriptions and extracted
features from descriptions. Based on the features, the CNN classified bitter melon leaf as
good or bad. The CNN based bitter melon prediction requires less computation time. But it
needs additional features to enhance the prediction accuracy.
3. Proposed Methodology
Here, the MME-DNN for crop yield prediction is described in detail. Initially, the
collected agriculture data (climate, weather and soil) is pre-processed by filling missing
values through multiple-imputation techniques. The pre-processed climate, weather and soil
data are processed by statistical models to know the variation of data from year-to-year. The
calculation of yield difference in a particular time period highlights the yield change that is
mainly due to changes of climate, weather and soil parameter. So the statistical model is
introduced to calculate the variation of climate, weather and soil parameter. The variation of
climate over the reproductive growth period of a crop is given as follows:
∆𝑪𝒕,𝒍,𝑺 = 𝑪𝒕,𝒍,𝑺 − 𝑪𝒕−𝟏:𝒕−𝟒,𝒍,𝑺 (1)
where, ∆𝑪𝒕,𝒍,𝑺 is the variation of climate in a year 𝒕 at a location𝒍which is at a height above
the sea level 𝑺. 𝑪𝒕,𝒍,𝑺 is the climate in a year 𝒕at a location𝒍 which is at a height above the sea
level 𝑺.𝑪𝒕−𝟏:𝒕−𝟒,𝒍,𝑺 is the climate in 𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑 and 𝒕 − 𝟒 years at a place 𝒍 which is
at a height above the sea level 𝑺.
The variation of weather over the reproductive growth of a crop is given as follows:
∆𝑾𝒕,𝒍,𝒂 = 𝑾𝒕,𝒍,𝒂 − 𝑾𝒕−𝟏:𝒕−𝟒,𝒍,𝒂 (2)
where, ∆𝑾𝒕,𝒍,𝒂 is the variation of weather in a year 𝒕 at a place 𝒍 with a air moisture level 𝒂.
𝑾𝒕,𝒍,𝒂 is the weather in a year 𝒕 at 𝒍 with 𝒂. 𝑾𝒕−𝟏:𝒕−𝟒,𝒍,𝒂 is the climate in𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑
and 𝒕 − 𝟒 years at a place 𝒍 with a air moisture level 𝒂.
The variation of soil parameters over a reproductive growth of a crop is given as follows:
∆𝑺𝒐𝒊𝒍𝒕,𝒍,𝒓 = 𝑺𝒐𝒊𝒍𝒕,𝒍,𝒓 − 𝑺𝒐𝒊𝒍𝒕−𝟏:𝒕−𝟒,𝒍,𝒓 (3)
where, ∆𝑺𝒐𝒊𝒍𝒕,𝒍,𝒓 is the variation of soil parameters in a year𝒕ata place 𝒍 that has water
retention capacity𝒓. 𝑺𝒐𝒊𝒍𝒕,𝒍,𝒓 is the soil parameters in a year 𝒕ata place 𝒍 that has water
retention capacity 𝒓. 𝑺𝒐𝒊𝒍𝒕−𝟏:𝒕−𝟒,𝒍,𝒓 is the soil parameter in 𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑 and 𝒕 −
𝟒 years ata place 𝒍that has water retention capacity 𝒓.
The variation of climate, weather and soil parameters is used in the climate, weather and soil
predictions. The climate, weather and soil predictions are ineluctable parameters for crop
yield prediction. But the climate, weather and soil parameters are unknown a priori. So,
neural network is applied to predict climate, weather and soil variables in a year 𝒕 that used
the historical climate, weather and soil data of previous years 𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑and 𝒕 − 𝟒.
Assume, 𝑿𝒍,𝒕
𝒄
represents the climate variable 𝒄 at place 𝒍 in year 𝒕. To predict the climate
variables in a year𝒕, the climate variables of previous years 𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑 and 𝒕 − 𝟒 and
their variations are trained in neural network. For each 𝒄, the neural network model explains
𝑿𝒍,𝒕
𝒄
as a reaction of four previous years at the same place which is denoted as
{𝑿𝒍,𝒕−𝟏
𝒄
, 𝑿𝒍,𝒕−𝟐
𝒄
, 𝑿𝒍,𝒕−𝟑
𝒄
, 𝑿𝒍,𝒕−𝟒
𝒄
}.
Consider, 𝑿𝒍,𝒕
𝒘
represents the weather variable 𝒘 at place 𝒍 in year 𝒕. The weather variables of
previous years 𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑 and 𝒕 − 𝟒 and their variations are trained in neural network

Vol. 28, No. 17, (2019), pp. 411-419
414
to predict the climate variables in a year 𝒕. For each 𝒘, the neural network explains the 𝑿𝒍,𝒕
𝒘
as
a reaction of four previous years at the same place which is denoted
as{𝑿𝒍,𝒕−𝟏
𝒘
, 𝑿𝒍,𝒕−𝟐
𝒘
, 𝑿𝒍,𝒕−𝟑
𝒘
, 𝑿𝒍,𝒕−𝟒
𝒘
}.
The soil variable 𝑺𝒐𝒊𝒍 at place 𝒍 in year 𝒕 is denoted as 𝑿𝒍,𝒕
𝑺𝒐𝒊𝒍
. The neural network is trained
with four previous years 𝒕 − 𝟏, 𝒕 − 𝟐, 𝒕 − 𝟑 and 𝒕 − 𝟒 soil variables and their variations to
predict the soil variables in a year 𝒕. For each 𝑺𝒐𝒊𝒍, the neural network model explains the
𝑿𝒍,𝒕
𝑺𝒐𝒊𝒍
as a reaction of four previous years at the same location: {𝑿𝒍,𝒕−𝟏
𝑺𝒐𝒊𝒍
, 𝑿𝒍,𝒕−𝟐
𝑺𝒐𝒊𝒍
, 𝑿𝒍,𝒕−𝟑
𝑺𝒐𝒊𝒍
, 𝑿𝒍,𝒕−𝟒
𝑺𝒐𝒊𝒍
}.
𝒙 denotes either climate, weather or soil variables and its variation is fed into the neural
network as input. The hidden layer of neural network is defined as tan-sigmoid transfer
function.
𝒇(𝒙) =
𝟐
𝟏+𝒆−𝟐𝒙 − 𝟏 (4)
where, 𝒙 is the input variable. Each climate, weather, soil variables and their variations has its
own weight values as 𝒘𝟏, 𝒘𝟐, … 𝒘𝒏 and adder function is used to perform weighted sum of
the inputs which is given as follows:
𝒖 = ∑ 𝒘𝒊𝒙𝒊
𝒏
𝒊=𝟏 (5)
where, 𝒏 is the number of variables.The output layer of neural network is described as:
𝒚 = 𝒇(∑ 𝒘𝒊𝒙𝒊 + 𝒃𝒊
𝒏
𝒊=𝟏 ) (6)
where, 𝒚 is the climate, weather and soil predictions, 𝒇(𝒙) is the transfer function, 𝒘𝒊 is the
weight values, 𝒙𝒊 is the climate, weather or soil variables and their variations and 𝒃𝒊 refers to
the bias value. The basic structure of neural network for climate, weather and soil prediction
is depicted in Figure 1.
Figure 1. Basic Neural Network Structure for Climate, weather and soil
prediction
3.1 Crop Yield Prediction
DNN is trained with predicted climate, weather and soil variables. One DNN is
trained for yield and one DNN is trained for check yield and their difference is the yield
difference. Maxout activation function is used for all neurons in the DNN except for the
output layer to introduce non-linearity into the output of a neuron. Residual shortcuts are used
in the DNN for every stacked hidden layer. It maps the identity and their results are included
with the outputs of the stacked layers.
DNN is similar to neural network whereas the difference is it consists of numerous
hidden layers between input and output layers. Here, the climate, weather and soil predictions
are given as input to the DNN. Each consequent layer assign weights to input parameters and
generates their output which is sent to the next layer. Finally at the output layer, the yield and
check yield is predicted. Figure 2 shows DNN structure for yield or check yield prediction.

Vol. 28, No. 17, (2019), pp. 411-419
415
Figure 2. DNN structure for Yield prediction
After the prediction of yield 𝒚 and check yield 𝒚𝒄 using DNN, the yield difference is
calculated as,
𝑽𝒂𝒓𝒊𝒂𝒏𝒄𝒆(𝒚𝒅) = 𝑽𝒂𝒓𝒊𝒂𝒏𝒄𝒆(𝒚 − 𝒚𝒄)
= 𝑽𝒂𝒓𝒊𝒂𝒏𝒄𝒆(𝒚) + 𝑽𝒂𝒓𝒊𝒂𝒏𝒄𝒆(𝒚𝒄) − 𝟐𝑪𝒐𝒗(𝒚,𝒚𝒄) (𝟕)
The yield difference is used to calculate the difference between yield and check yield which
helps to know the accuracy of the yield. Thus the MME-DNN predicts the crop yield using
the climate, weather and soil variables.
MME-DNN based Crop Yield Prediction Algorithm
Step 1:Collect climate, weather and soil parameters
Step 2:Calculate the variation of climate, weather and soil over the reproductive growth of a
crop using Eq. (1-3).
Step 3: Predict the climate, weather and soil variables by applying DNN
Step 4: Predict the yield and check yield using Eq. (4-6)
Step5: Calculate the difference between yield and check yield to know the accuracy of yield
prediction.
4. Result and Discussion
Here, the efficiency of MME-DNN and DNN based crop yield prediction methods
are evaluated based on its accuracy, precision, recall and F-measure value. For the
experimental purpose, climate data are collected from http://www.ccafs-
climate.org/climatewizard/, crop data are collected from
https://data.world/thatzprem/agriculture-indiaand soil data are collected from
https://data.gov.in/search/site?query=soil. From the crop, climate and soil data 7000 data are
used for training and 30,000 data are used for testing.
4.1 Accuracy
It is fraction of correct crop yield prediction over the total number of instances evaluated.
𝑨𝒄𝒄𝒖𝒓𝒂𝒄𝒚
=
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆 + 𝑻𝒓𝒖𝒆 𝑵𝒆𝒈𝒂𝒕𝒊𝒗𝒆
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆 + 𝑻𝒓𝒖𝒆 𝑵𝒆𝒈𝒂𝒕𝒊𝒗𝒆 + 𝑭𝒂𝒍𝒔𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆 + 𝑭𝒂𝒍𝒔𝒆 𝑵𝒆𝒈𝒂𝒕𝒊𝒗𝒆
The accuracy of DNN and MME-DNN based crop yield prediction method for five different
crops is shown in Table 1.
Crop yield prediction method Banana Groundnut Wheat Sugarcane Maize
DNN 0.88 0.91 0.89 0.88 0.88
MME-DNN 0.90 0.92 0.90 0.91 0.92

Vol. 28, No. 17, (2019), pp. 411-419
416
Table.2 Comparison of Accuracy
The testing performance between DNN and MME-DNN based crop yield prediction
method for five different crops in terms of accuracy is shown in accuracy. The accuracy of
MME-DNN for banana is 2.27%, for groundnut is 1.1%, for wheat is 1.12%, for sugarcane is
3.41% and for maize is 4.55% which is greater than DNN based crop yield prediction
method. From Figure 3, it is proved that the MME-DNN based crop yield prediction method
has high accuracy than DNN based crop yield prediction for five different crops.
Figure.3 Evaluation of Accuracy
4.2 Precision
It is calculated based on crop yield prediction at True Positive and False Positive rates. It is
computed as
𝑷𝒓𝒆𝒄𝒊𝒔𝒊𝒐𝒏 =
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆 + 𝑭𝒂𝒔𝒍𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆
The precision of DNN and MME-DNN based crop yield prediction method for five different
DNN 0.51 0.91 0.85 0.78 0.57
MME-DNN 0.56 0.92 0.86 0.86 0.60
Table.3 Comparison of Precision
Figure.4 Evaluation of Precision

Vol. 28, No. 17, (2019), pp. 411-419
417
The obtained values for the precision rate using the proposed method MME-DNN are
greater than is compared with the proposed method MME-DNN method are greater than
(9.8%, 1.1%, 1.18%, 10.25%and 5.26%) the existing method DNN for the different crops for
banana, groundnut, wheat, sugarcane and maize crop respectively. From Figure 4, it is clear
that the MME-DNN has better precision than DNN-based crop yield prediction for different
crops.
4.3 Recall
It is calculated based on the crop yield prediction at true positive and false negative rates.
𝑹𝒆𝒄𝒂𝒍𝒍 =
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆 + 𝑭𝒂𝒍𝒔𝒆 𝑵𝒆𝒈𝒂𝒕𝒊𝒗𝒆
The recall of DNN and MME-DNN based crop yield prediction method for five different
DNN 0.89 0.9 0.91 0.9 0.9
MME-DNN 0.93 0.92 0.92 0.93 0.92
Table.4 Comparison of Recall
Figure.5 Evaluation of Recall
The recall for DNN and MME-DNN based crop yield prediction methods using five
different crops is shown in Figure 5. This analysis indicates that the proposed MME-DNN
method achieves higher recall than the DNN for predicting five different crop yields. For
instance, the recall of MME-DNN is 4.49%, 2.22%, 1.1%, 3.33% and 2.22% for different
crops which is higher than the DNN method for banana, groundnut, wheat, sugarcane and
maize, respectively.
4.4 F-measure
F-measure is mean value of precision and recall. It is calculated as,
𝑭 − 𝒎𝒆𝒂𝒔𝒖𝒓𝒆 = 𝟐 ∙ (
𝑷𝒓𝒆𝒄𝒊𝒔𝒊𝒐𝒏 ∙ 𝑹𝒆𝒄𝒂𝒍𝒍
𝑷𝒓𝒆𝒄𝒊𝒔𝒊𝒐𝒏 + 𝑹𝒆𝒄𝒂𝒍𝒍
)
The F-measure of DNN and MME-DNN based crop yield prediction method for five different
crops is Table 5.

Vol. 28, No. 17, (2019), pp. 411-419
418
DNN 0.78 0.91 0.88 0.86 0.80
MME-DNN 0.81 0.92 0.90 0.90 0.82
Table.5 Comparison of F-measure
Figure.6 Comparison of F-measure
F-measure value of DNN and MME-DNN based crop yield prediction methods for
five different crops is shown in Figure 6. The F-measure of MME-DNN for banana is 3.85%,
for groundnut is 1.1%, for wheat is 2.27%, for sugarcane is 4.65% and for maize is 2.5%
which is greater than DNN based crop yield prediction method. Hence, it is clear that the
MME-DNN has high f-measure than DNN based crop yield prediction method for banana,
5. Conclusion
In this paper, a MME-DNN is proposed for crop yield prediction. Based on proposed
MME-DNN, the crop yield is predicted through climate, weather and soil variables and their
variations. The variation of climate, weather and soil variables is predicted by using statistical
model. A neural network is used to predict the climate, weather and soil variables and the
predicted variables are processed in DNN for yield prediction. The experimental results show
that the proposed MME-DNN has high accuracy, precision, recall and F-measure than DNN
for five different crops.
References
[1] Bhanose, S. S., Bogawar, K. A., Dhotre, A. G., & Gaidhani, B. R. (2016). Crop and Yield Prediction
Model. International Journal of Advance Scientific Research and Engineering Trends, 1(1), 23-28.
[2] Awad, M. M. (2019). An innovative intelligent system based on remote sensing and mathematical
models for improving crop yield estimation. Information Processing in Agriculture.
[3] Sujatha, R., & Isakki, P. (2016, January). A study on crop yield forecasting using classification
techniques. In 2016 International Conference on Computing Technologies and Intelligent Data
Engineering (ICCTIDE'16) (pp. 1-4). IEEE.
[4] Gandge, Y. (2017, December). A study on various data mining techniques for crop yield prediction.
In 2017 International Conference on Electrical, Electronics, Communication, Computer, and
Optimization Techniques (ICEECCOT) (pp. 420-423). IEEE.
[5] Khaki, S., & Wang, L. (2019). Crop yield prediction using deep neural networks. Frontiers in plant
science, 10.
[6] Shastry, A., Sanjay, H. A., & Hegde, M. (2015, June). A parameter based ANFIS model for crop yield
prediction. In 2015 IEEE International Advance Computing Conference (IACC) (pp. 253-257). IEEE.
[7] Huang, X., Huang, G., Yu, C., Ni, S., & Yu, L. (2017). A multiple crop model ensemble for improving
broad-scale yield prediction using Bayesian model averaging. Field Crops Research, 211, 114-124.
[8] Mohan, P., & Patil, K. K. (2017). Weather and Crop Prediction Using Modified Self Organizing Map for
Mysore Region. International Journal of Intelligent Engineering & Systems (IJIES), 11(2), 192-199.
[9] Patel, H., & Patel, D. (2017). Crop yield prediction using rough set theory. International Journal of
Engineering and Technology (IJET), 9(03), 2505-2513.

Vol. 28, No. 17, (2019), pp. 411-419
419
[10] Shah, A., Dubey, A., Hemnani, V., Gala, D., & Kalbande, D. R. (2018). Smart Farming System: Crop
Yield Prediction Using Regression Techniques. In Proceedings of International Conference on Wireless
Communication (pp. 49-56). Springer, Singapore.
[11] Verma, A., Jatain, A., & Bajaj, S. (2018). Crop yield prediction of wheat using Fuzzy C Means
clustering and neural network. International Journal of Applied Engineering Research, 13(11), 9816-
9821.
[12] Villanueva, M. B., & Salenga, M. L. M. (2018). Bitter Melon Crop Yield Prediction using Machine
Learning Algorithm. International Journal of Advanced Computer Science and Applications (IJACSA), 9
(3), 1-6.

Crop yield prediction using deep learning.pdf

More Related Content

Similar to Crop yield prediction using deep learning.pdf (20)

Recently uploaded (20)

Crop yield prediction using deep learning.pdf