Welcome to the Crop Recommendation and Yield Prediction project! This project aims to provide valuable insights to farmers by recommending the best crops to cultivate based on soil and weather conditions and predicting crop yields to optimize agricultural practices.
Agriculture plays a vital role in the economy, and optimizing crop selection and yield prediction can significantly enhance productivity and sustainability. This project leverages machine learning techniques to recommend suitable crops and predict their yields based on weather patterns and soil conditions.
Data preparation is the most challenging task for building this application. Our aim is to generate crop recommendations and yield predictions for farmers based on parameters like temperature, humidity, rainfall, and soil conditions such as N, P, K, and pH using Machine Learning.
The most challenging part is finding data to train ML algorithms. Although there is vast data available on the internet, it is often random, unorganized, and isolated to each attribute. We need to use the available data from various sources and organize it in the right format to train an ML algorithm that is diverse enough to recommend crops or predict yield in various weather and soil conditions, instead of confining it to a particular region limited by data availability.
The objective is to prepare data to train machine learning models for recommending crops to farmers and predicting crop yields based on weather and soil parameters.
The data is sourced from various datasets available on Kaggle:
- Crop Recommendation Dataset: Contains soil parameters and recommended crops.
- Indian Agriculture Crop Production Dataset: Contains crop production data.
- District Coordinates Dataset: Contains geographic coordinates of districts.
- Soil Data: Contains soil information.
+-----------------------------------------------------------+
| 🚀 Start |
+-----------------------------------------------------------+
|
v
+-----------------------------------------------------------+
| 📥 Load Datasets |
|-----------------------------------------------------------|
| - Load crop recommendation data |
| - Load crop production data |
| - Load district coordinates |
| - Load soil data |
+-----------------------------------------------------------+
|
v
+-----------------------------------------------------------+
| 🔄 Merge Datasets |
|-----------------------------------------------------------|
| - Merge crop and production data on 'District' |
| - Merge with district coordinates on 'District' |
| - Merge with soil data on 'District' |
+-----------------------------------------------------------+
|
v
+-----------------------------------------------------------+
| ☁️ Fetch Weather Data |
|-----------------------------------------------------------|
| - Define API function to fetch weather data |
| - Iterate through merged data and fetch weather data |
| - Process and merge weather data |
+-----------------------------------------------------------+
|
v
+-----------------------------------------------------------+
| 🛠️ Final Data Preparation |
|-----------------------------------------------------------|
| - Process combined data |
| - Handle missing values, normalize features |
| - Save final combined data to CSV |
+-----------------------------------------------------------+
|
v
+-----------------------------------------------------------+
| 🏁 End |
+-----------------------------------------------------------+
The crop recommendation model is designed to suggest the most suitable crops for cultivation based on various soil parameters and weather conditions. By leveraging machine learning techniques, the model analyzes the relationship between input features and the recommended crops to provide accurate and data-driven recommendations to farmers.
- n_estimators: 100
- max_depth: 10
- min_samples_split: 2
- min_samples_leaf: 1
precision recall f1-score support
0 1.00 1.00 1.00 3
1 1.00 1.00 1.00 12
2 1.00 1.00 1.00 13
3 1.00 1.00 1.00 7
4 1.00 1.00 1.00 15
accuracy 1.00 50
macro avg 1.00 1.00 1.00 50
weighted avg 1.00 1.00 1.00 50
.:. Random Forest: 100.000% .:.
Note : Here the model had great score with test data also but the 100% accuracy is due to less data availability.
The model will perform better when retrained wwith large amount of data. But limited availability of data might compromise it capability
The yield prediction model is designed to estimate the agricultural yield of crops based on various input features such as soil properties, weather conditions, and historical production data. This model uses the Random Forest Classifier to analyze complex relationships between the input parameters and yield outcomes, providing accurate yield predictions to help farmers and agricultural planners make informed decisions.
- n_estimators: 568
- min_samples_split: 16
- min_samples_leaf: 7
- max_depth: 3
- RMSE for Random Forest: 0.5483
POST: https://karshaka-ml-models.onrender.com/recomendCrop
{
"N": 20,
"P": 20,
"K": 40,
"temperature": 30,
"humidity": 50,
"ph": 8,
"rainfall": 120
}- first_crop: The most suitable crop (string)
- second_crop: The second most suitable crop (string)
import requests
url = "https://karshaka-ml-models.onrender.com/recomendCrop"
payload = {
"N": , # fill with your value
"P": , # fill with your value
"K": , # fill with your value
"temperature": , # fill with your value
"humidity": , # fill with your value
"ph": , # fill with your value
"rainfall": # fill with your value
}
response = requests.post(url, json=payload)
if response.status_code == 200:
data = response.json()
print("First Crop:", data['first_crop'])
print("Second Crop:", data['second_crop'])
else:
print("Error:", response.status_code, response.text)const url = "https://karshaka-ml-models.onrender.com/recomendCrop";
const payload = {
"N": , // fill with your value
"P": , // fill with your value
"K": , // fill with your value
"temperature": , // fill with your value
"humidity": , // fill with your value
"ph": , // fill with your value
"rainfall": // fill with your value
};
fetch(url, {
method: "POST",
headers: {
"Content-Type": "application/json"
},
body: JSON.stringify(payload)
})
.then(response => response.json())
.then(data => {
console.log("First Crop:", data.first_crop);
console.log("Second Crop:", data.second_crop);
})
.catch(error => {
console.error("Error:", error);
});POST: https://karshaka-ml-models.onrender.com/predictYeild
This endpoint predicts the yield, crop can make based on soil and weather parameters.
{
"Crop": "rice",
"Avg_Nitrogen": 80,
"Avg_Phosphorous": 40,
"Avg_Potassium": 30,
"pH": 6,
"temperature": 30,
"humidity": 80,
"rainfall": 220
}- prediction: The predicted yield of the crop (float)
import requests
url = "https://karshaka-ml-models.onrender.com/predictYeild"
payload = {
"Crop": , # fill with your value
"Avg_Nitrogen": , # fill with your value
"Avg_Phosphorous": , # fill with your value
"Avg_Potassium": , # fill with your value
"pH": , # fill with your value
"temperature": , # fill with your value
"humidity": , # fill with your value
"rainfall": # fill with your value
}
response = requests.post(url, json=payload)
if response.status_code == 200:
data = response.json()
print("Predicted Yield:", data['prediction'])
else:
print("Error:", response.status_code, response.text)const url = "https://karshaka-ml-models.onrender.com/predictYeild";
const payload = {
"Crop": , // fill with your value
"Avg_Nitrogen": , // fill with your value
"Avg_Phosphorous": , // fill with your value
"Avg_Potassium": , // fill with your value
"pH": , // fill with your value
"temperature": , // fill with your value
"humidity": , // fill with your value
"rainfall": // fill with your value
};
fetch(url, {
method: "POST",
headers: {
"Content-Type": "application/json"
},
body: JSON.stringify(payload)
})
.then(response => response.json())
.then(data => {
console.log("Predicted Yield:", data.prediction);
})
.catch(error => {
console.error("Error:", error);
});