Jupyter Notebook · Machine Learning
Crop Recommendation Model
Supervised classification that recommends the best crop from soil and climate inputs (N, P, K, temperature, humidity, pH, rainfall), benchmarking six models and reaching about 99% accuracy.
In [1]:
# Importing libraries
from __future__ import print_function
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.metrics import classification_report
from sklearn import metrics
from sklearn import tree
import warnings
warnings.filterwarnings('ignore')In [2]:
df = pd.read_csv('../Data-processed/crop-recommendation.csv')In [3]:
df.head()| N | P | K | temperature | humidity | ph | rainfall | label | |
|---|---|---|---|---|---|---|---|---|
| 0 | 90 | 42 | 43 | 20.879744 | 82.002744 | 6.502985 | 202.935536 | rice |
| 1 | 85 | 58 | 41 | 21.770462 | 80.319644 | 7.038096 | 226.655537 | rice |
| 2 | 60 | 55 | 44 | 23.004459 | 82.320763 | 7.840207 | 263.964248 | rice |
| 3 | 74 | 35 | 40 | 26.491096 | 80.158363 | 6.980401 | 242.864034 | rice |
| 4 | 78 | 42 | 42 | 20.130175 | 81.604873 | 7.628473 | 262.717340 | rice |
In [4]:
df.tail()| N | P | K | temperature | humidity | ph | rainfall | label | |
|---|---|---|---|---|---|---|---|---|
| 2195 | 107 | 34 | 32 | 26.774637 | 66.413269 | 6.780064 | 177.774507 | coffee |
| 2196 | 99 | 15 | 27 | 27.417112 | 56.636362 | 6.086922 | 127.924610 | coffee |
| 2197 | 118 | 33 | 30 | 24.131797 | 67.225123 | 6.362608 | 173.322839 | coffee |
| 2198 | 117 | 32 | 34 | 26.272418 | 52.127394 | 6.758793 | 127.175293 | coffee |
| 2199 | 104 | 18 | 30 | 23.603016 | 60.396475 | 6.779833 | 140.937041 | coffee |
In [5]:
df.size17600
In [6]:
df.shape(2200, 8)
In [7]:
df.columnsIndex(['N', 'P', 'K', 'temperature', 'humidity', 'ph', 'rainfall', 'label'], dtype='object')
In [8]:
df['label'].unique()array(['rice', 'maize', 'chickpea', 'kidneybeans', 'pigeonpeas',
'mothbeans', 'mungbean', 'blackgram', 'lentil', 'pomegranate',
'banana', 'mango', 'grapes', 'watermelon', 'muskmelon', 'apple',
'orange', 'papaya', 'coconut', 'cotton', 'jute', 'coffee'],
dtype=object)In [9]:
df.dtypesN int64 P int64 K int64 temperature float64 humidity float64 ph float64 rainfall float64 label object dtype: object
In [10]:
df['label'].value_counts()muskmelon 100 kidneybeans 100 papaya 100 pigeonpeas 100 blackgram 100 cotton 100 mothbeans 100 mungbean 100 watermelon 100 orange 100 mango 100 banana 100 rice 100 pomegranate 100 chickpea 100 apple 100 jute 100 grapes 100 lentil 100 coffee 100 maize 100 coconut 100 Name: label, dtype: int64
In [11]:
sns.heatmap(df.corr(),annot=True)
Seperating features and target label
In [12]:
features = df[['N', 'P','K','temperature', 'humidity', 'ph', 'rainfall']]
target = df['label']
#features = df[['temperature', 'humidity', 'ph', 'rainfall']]
labels = df['label']In [13]:
# Initialzing empty lists to append all model's name and corresponding name
acc = []
model = []In [14]:
# Splitting into train and test data
from sklearn.model_selection import train_test_split
Xtrain, Xtest, Ytrain, Ytest = train_test_split(features,target,test_size = 0.2,random_state =2)Decision Tree
In [15]:
from sklearn.tree import DecisionTreeClassifier
DecisionTree = DecisionTreeClassifier(criterion="entropy",random_state=2,max_depth=5)
DecisionTree.fit(Xtrain,Ytrain)
predicted_values = DecisionTree.predict(Xtest)
x = metrics.accuracy_score(Ytest, predicted_values)
acc.append(x)
model.append('Decision Tree')
print("DecisionTrees's Accuracy is: ", x*100)
print(classification_report(Ytest,predicted_values))DecisionTrees's Accuracy is: 90.0
precision recall f1-score support
apple 1.00 1.00 1.00 13
banana 1.00 1.00 1.00 17
blackgram 0.59 1.00 0.74 16
chickpea 1.00 1.00 1.00 21
coconut 0.91 1.00 0.95 21
coffee 1.00 1.00 1.00 22
cotton 1.00 1.00 1.00 20
grapes 1.00 1.00 1.00 18
jute 0.74 0.93 0.83 28
kidneybeans 0.00 0.00 0.00 14
lentil 0.68 1.00 0.81 23
maize 1.00 1.00 1.00 21
mango 1.00 1.00 1.00 26
mothbeans 0.00 0.00 0.00 19
mungbean 1.00 1.00 1.00 24
muskmelon 1.00 1.00 1.00 23
orange 1.00 1.00 1.00 29
papaya 1.00 0.84 0.91 19
pigeonpeas 0.62 1.00 0.77 18
pomegranate 1.00 1.00 1.00 17
rice 1.00 0.62 0.77 16
watermelon 1.00 1.00 1.00 15
accuracy 0.90 440
macro avg 0.84 0.88 0.85 440
weighted avg 0.86 0.90 0.87 440
In [16]:
from sklearn.model_selection import cross_val_scoreIn [17]:
# Cross validation score (Decision Tree)
score = cross_val_score(DecisionTree, features, target,cv=5)In [18]:
scorearray([0.93636364, 0.90909091, 0.91818182, 0.87045455, 0.93636364])
Saving trained Decision Tree model
In [19]:
import pickle
# Dump the trained Naive Bayes classifier with Pickle
DT_pkl_filename = '../models/DecisionTree.pkl'
# Open the file to save as pkl file
DT_Model_pkl = open(DT_pkl_filename, 'wb')
pickle.dump(DecisionTree, DT_Model_pkl)
# Close the pickle instances
DT_Model_pkl.close()Guassian Naive Bayes
In [20]:
from sklearn.naive_bayes import GaussianNB
NaiveBayes = GaussianNB()
NaiveBayes.fit(Xtrain,Ytrain)
predicted_values = NaiveBayes.predict(Xtest)
x = metrics.accuracy_score(Ytest, predicted_values)
acc.append(x)
model.append('Naive Bayes')
print("Naive Bayes's Accuracy is: ", x)
print(classification_report(Ytest,predicted_values))Naive Bayes's Accuracy is: 0.990909090909091
precision recall f1-score support
apple 1.00 1.00 1.00 13
banana 1.00 1.00 1.00 17
blackgram 1.00 1.00 1.00 16
chickpea 1.00 1.00 1.00 21
coconut 1.00 1.00 1.00 21
coffee 1.00 1.00 1.00 22
cotton 1.00 1.00 1.00 20
grapes 1.00 1.00 1.00 18
jute 0.88 1.00 0.93 28
kidneybeans 1.00 1.00 1.00 14
lentil 1.00 1.00 1.00 23
maize 1.00 1.00 1.00 21
mango 1.00 1.00 1.00 26
mothbeans 1.00 1.00 1.00 19
mungbean 1.00 1.00 1.00 24
muskmelon 1.00 1.00 1.00 23
orange 1.00 1.00 1.00 29
papaya 1.00 1.00 1.00 19
pigeonpeas 1.00 1.00 1.00 18
pomegranate 1.00 1.00 1.00 17
rice 1.00 0.75 0.86 16
watermelon 1.00 1.00 1.00 15
accuracy 0.99 440
macro avg 0.99 0.99 0.99 440
weighted avg 0.99 0.99 0.99 440
In [21]:
# Cross validation score (NaiveBayes)
score = cross_val_score(NaiveBayes,features,target,cv=5)
scorearray([0.99772727, 0.99545455, 0.99545455, 0.99545455, 0.99090909])
Saving trained Guassian Naive Bayes model
In [22]:
import pickle
# Dump the trained Naive Bayes classifier with Pickle
NB_pkl_filename = '../models/NBClassifier.pkl'
# Open the file to save as pkl file
NB_Model_pkl = open(NB_pkl_filename, 'wb')
pickle.dump(NaiveBayes, NB_Model_pkl)
# Close the pickle instances
NB_Model_pkl.close()Support Vector Machine (SVM)
In [23]:
from sklearn.svm import SVC
# data normalization with sklearn
from sklearn.preprocessing import MinMaxScaler
# fit scaler on training data
norm = MinMaxScaler().fit(Xtrain)
X_train_norm = norm.transform(Xtrain)
# transform testing dataabs
X_test_norm = norm.transform(Xtest)
SVM = SVC(kernel='poly', degree=3, C=1)
SVM.fit(X_train_norm,Ytrain)
predicted_values = SVM.predict(X_test_norm)
x = metrics.accuracy_score(Ytest, predicted_values)
acc.append(x)
model.append('SVM')
print("SVM's Accuracy is: ", x)
print(classification_report(Ytest,predicted_values))SVM's Accuracy is: 0.9795454545454545
precision recall f1-score support
apple 1.00 1.00 1.00 13
banana 1.00 1.00 1.00 17
blackgram 1.00 1.00 1.00 16
chickpea 1.00 1.00 1.00 21
coconut 1.00 1.00 1.00 21
coffee 1.00 0.95 0.98 22
cotton 0.95 1.00 0.98 20
grapes 1.00 1.00 1.00 18
jute 0.83 0.89 0.86 28
kidneybeans 1.00 1.00 1.00 14
lentil 1.00 1.00 1.00 23
maize 1.00 0.95 0.98 21
mango 1.00 1.00 1.00 26
mothbeans 1.00 1.00 1.00 19
mungbean 1.00 1.00 1.00 24
muskmelon 1.00 1.00 1.00 23
orange 1.00 1.00 1.00 29
papaya 1.00 1.00 1.00 19
pigeonpeas 1.00 1.00 1.00 18
pomegranate 1.00 1.00 1.00 17
rice 0.80 0.75 0.77 16
watermelon 1.00 1.00 1.00 15
accuracy 0.98 440
macro avg 0.98 0.98 0.98 440
weighted avg 0.98 0.98 0.98 440
In [24]:
# Cross validation score (SVM)
score = cross_val_score(SVM,features,target,cv=5)
scorearray([0.97954545, 0.975 , 0.98863636, 0.98863636, 0.98181818])
In [25]:
#Saving trained SVM modelIn [26]:
import pickle
# Dump the trained SVM classifier with Pickle
SVM_pkl_filename = '../models/SVMClassifier.pkl'
# Open the file to save as pkl file
SVM_Model_pkl = open(SVM_pkl_filename, 'wb')
pickle.dump(SVM, SVM_Model_pkl)
# Close the pickle instances
SVM_Model_pkl.close()Logistic Regression
In [27]:
from sklearn.linear_model import LogisticRegression
LogReg = LogisticRegression(random_state=2)
LogReg.fit(Xtrain,Ytrain)
predicted_values = LogReg.predict(Xtest)
x = metrics.accuracy_score(Ytest, predicted_values)
acc.append(x)
model.append('Logistic Regression')
print("Logistic Regression's Accuracy is: ", x)
print(classification_report(Ytest,predicted_values))Logistic Regression's Accuracy is: 0.9522727272727273
precision recall f1-score support
apple 1.00 1.00 1.00 13
banana 1.00 1.00 1.00 17
blackgram 0.86 0.75 0.80 16
chickpea 1.00 1.00 1.00 21
coconut 1.00 1.00 1.00 21
coffee 1.00 1.00 1.00 22
cotton 0.86 0.90 0.88 20
grapes 1.00 1.00 1.00 18
jute 0.84 0.93 0.88 28
kidneybeans 1.00 1.00 1.00 14
lentil 0.88 1.00 0.94 23
maize 0.90 0.86 0.88 21
mango 0.96 1.00 0.98 26
mothbeans 0.84 0.84 0.84 19
mungbean 1.00 0.96 0.98 24
muskmelon 1.00 1.00 1.00 23
orange 1.00 1.00 1.00 29
papaya 1.00 0.95 0.97 19
pigeonpeas 1.00 1.00 1.00 18
pomegranate 1.00 1.00 1.00 17
rice 0.85 0.69 0.76 16
watermelon 1.00 1.00 1.00 15
accuracy 0.95 440
macro avg 0.95 0.95 0.95 440
weighted avg 0.95 0.95 0.95 440
In [28]:
# Cross validation score (Logistic Regression)
score = cross_val_score(LogReg,features,target,cv=5)
scorearray([0.95 , 0.96590909, 0.94772727, 0.96590909, 0.94318182])
Saving trained Logistic Regression model
In [29]:
import pickle
# Dump the trained Naive Bayes classifier with Pickle
LR_pkl_filename = '../models/LogisticRegression.pkl'
# Open the file to save as pkl file
LR_Model_pkl = open(DT_pkl_filename, 'wb')
pickle.dump(LogReg, LR_Model_pkl)
# Close the pickle instances
LR_Model_pkl.close()Random Forest
In [30]:
from sklearn.ensemble import RandomForestClassifier
RF = RandomForestClassifier(n_estimators=20, random_state=0)
RF.fit(Xtrain,Ytrain)
predicted_values = RF.predict(Xtest)
x = metrics.accuracy_score(Ytest, predicted_values)
acc.append(x)
model.append('RF')
print("RF's Accuracy is: ", x)
print(classification_report(Ytest,predicted_values))RF's Accuracy is: 0.990909090909091
precision recall f1-score support
apple 1.00 1.00 1.00 13
banana 1.00 1.00 1.00 17
blackgram 0.94 1.00 0.97 16
chickpea 1.00 1.00 1.00 21
coconut 1.00 1.00 1.00 21
coffee 1.00 1.00 1.00 22
cotton 1.00 1.00 1.00 20
grapes 1.00 1.00 1.00 18
jute 0.90 1.00 0.95 28
kidneybeans 1.00 1.00 1.00 14
lentil 1.00 1.00 1.00 23
maize 1.00 1.00 1.00 21
mango 1.00 1.00 1.00 26
mothbeans 1.00 0.95 0.97 19
mungbean 1.00 1.00 1.00 24
muskmelon 1.00 1.00 1.00 23
orange 1.00 1.00 1.00 29
papaya 1.00 1.00 1.00 19
pigeonpeas 1.00 1.00 1.00 18
pomegranate 1.00 1.00 1.00 17
rice 1.00 0.81 0.90 16
watermelon 1.00 1.00 1.00 15
accuracy 0.99 440
macro avg 0.99 0.99 0.99 440
weighted avg 0.99 0.99 0.99 440
In [31]:
# Cross validation score (Random Forest)
score = cross_val_score(RF,features,target,cv=5)
scorearray([0.99772727, 0.99545455, 0.99772727, 0.99318182, 0.98863636])
Saving trained Random Forest model
In [32]:
import pickle
# Dump the trained Naive Bayes classifier with Pickle
RF_pkl_filename = '../models/RandomForest.pkl'
# Open the file to save as pkl file
RF_Model_pkl = open(RF_pkl_filename, 'wb')
pickle.dump(RF, RF_Model_pkl)
# Close the pickle instances
RF_Model_pkl.close()XGBoost
In [33]:
import xgboost as xgb
XB = xgb.XGBClassifier()
XB.fit(Xtrain,Ytrain)
predicted_values = XB.predict(Xtest)
x = metrics.accuracy_score(Ytest, predicted_values)
acc.append(x)
model.append('XGBoost')
print("XGBoost's Accuracy is: ", x)
print(classification_report(Ytest,predicted_values))[14:16:03] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior.
XGBoost's Accuracy is: 0.9931818181818182
precision recall f1-score support
apple 1.00 1.00 1.00 13
banana 1.00 1.00 1.00 17
blackgram 1.00 1.00 1.00 16
chickpea 1.00 1.00 1.00 21
coconut 1.00 1.00 1.00 21
coffee 0.96 1.00 0.98 22
cotton 1.00 1.00 1.00 20
grapes 1.00 1.00 1.00 18
jute 1.00 0.93 0.96 28
kidneybeans 1.00 1.00 1.00 14
lentil 0.96 1.00 0.98 23
maize 1.00 1.00 1.00 21
mango 1.00 1.00 1.00 26
mothbeans 1.00 0.95 0.97 19
mungbean 1.00 1.00 1.00 24
muskmelon 1.00 1.00 1.00 23
orange 1.00 1.00 1.00 29
papaya 1.00 1.00 1.00 19
pigeonpeas 1.00 1.00 1.00 18
pomegranate 1.00 1.00 1.00 17
rice 0.94 1.00 0.97 16
watermelon 1.00 1.00 1.00 15
accuracy 0.99 440
macro avg 0.99 0.99 0.99 440
weighted avg 0.99 0.99 0.99 440
In [34]:
# Cross validation score (XGBoost)
score = cross_val_score(XB,features,target,cv=5)
score[08:54:44] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior. [08:54:45] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior. [08:54:46] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior. [08:54:47] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior. [08:54:48] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior.
array([0.99318182, 0.99318182, 0.99318182, 0.99090909, 0.99090909])
Saving trained XGBoost model
In [35]:
import pickle
# Dump the trained Naive Bayes classifier with Pickle
XB_pkl_filename = '../models/XGBoost.pkl'
# Open the file to save as pkl file
XB_Model_pkl = open(XB_pkl_filename, 'wb')
pickle.dump(XB, XB_Model_pkl)
# Close the pickle instances
XB_Model_pkl.close()Accuracy Comparison
In [36]:
plt.figure(figsize=[10,5],dpi = 100)
plt.title('Accuracy Comparison')
plt.xlabel('Accuracy')
plt.ylabel('Algorithm')
sns.barplot(x = acc,y = model,palette='dark')
In [37]:
accuracy_models = dict(zip(model, acc))
for k, v in accuracy_models.items():
print (k, '-->', v)Decision Tree --> 0.9 Naive Bayes --> 0.990909090909091 SVM --> 0.9795454545454545 Logistic Regression --> 0.9522727272727273 RF --> 0.990909090909091 XGBoost --> 0.9931818181818182
Making a prediction
In [38]:
data = np.array([[104,18, 30, 23.603016, 60.3, 6.7, 140.91]])
prediction = RF.predict(data)
print(prediction)['coffee']
In [39]:
data = np.array([[83, 45, 60, 28, 70.3, 7.0, 150.9]])
prediction = RF.predict(data)
print(prediction)['jute']