Skip to content

PThie/Prediction-Agriculture

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Project: Predictive Modelling for Crops

Objective

  • Find the feature (chemical element) that predicts the crop type the best

Data

  • Various preferred concentration combinations of chemical elements for different crop types
  • Source: DataCamp

Strategy

  • Multi-class classification to group crops given the conditions under which they thrive
  • Apply logistic regression for each feature (concentration) to identify which soil characteristic determines the crop type the best

Describing the data

Descriptive statistics

Nitrogen
concentration
Phosphorous
concentration
Potassium
concentration
PH value
mean 50.6 53.4 48.1 6.5
std 36.9 33.0 50.6 0.8
min 0.0 5.0 5.0 3.5
25% 21.0 28.0 20.0 6.0
50% 37.0 51.0 32.0 6.4
75% 84.2 68.0 49.0 6.9
max 140.0 145.0 205.0 9.9

Distribution of concentration values by crop

Scatter plot of nitrogen concentration Scatter plot of phosphorous concentration
Scatter plot of potassium concentration Scatter plot of PH values
  • Nitrogen concentration: Median concentrations for many crops centers around 20 but few outliers with larger concentration values.
  • Phosphorous concentration: Broader values than for Nitrogen. Most crops prefer concentrations below 100, except Grapes and Apples.
  • Potassium concentration: Most crops prefer lower concentration values (between 25 and 50) but Apples and Grapes show extreme values around 200.
  • PH values are mostly between 6 and 7 which makes sense as crops cannot flourish under too alkaline or acidic conditions. However, Mothbeans show a broad distribution (values between 4 and 10).

Pairwise-correlations between concentration types

Nitrogen
concentration
Phosphorous
concentration
Potassium
concentration
PH value
Nitrogen
concentration
1.00 -0.23 -0.14 0.10
Phosphorous
concentration
-0.23 1.00 0.74 -0.14
Potassium
concentration
-0.14 0.74 1.00 -0.17
PH value 0.10 -0.14 -0.17 1.00
  • Phosphorous and Potassium are strongly positively correlated
  • Nitrogen and Phosphorous have a weak to moderate negative correlation

Key findings

Prediction outcome: F1 score

concentration F1 score
Nitrogen
concentration
0.104057
Phosphorous
concentration
0.182618
Potassium
concentration
0.260034
PH value 0.076458

Summary

  • Best predictor for the crop type: Potassium (F1 Score = 0.26) followed by Phosphorous (F1 score = 0.18)
  • That Phosphorous follows Potassium quite closely and is also a (relatively) good predictor might be suspected from the correlation table above. Phosphorous and Potassium are strongly positively correlated.
  • PH value is a bad predictor for the crops.

About

Predictive Modelling for Agriculture Crops

Topics

Resources

Stars

Watchers

Forks