ML---Soil-Quality

A machine learning model to predict soil fertility based on elemental soil analysis

Requirments

Requirments

Data

Data Source

Raw Data is published on kaggle

• Original Dataset
• Modified Dataset

Atrributes

• N - ratio of Nitrogen (NH4+) content in soil
• P - ratio of Phosphorous (P) content in soil
• K - ratio of Potassium (K) content in soil
• ph - soil acidity (pH)
• ec - electrical conductivity
• oc - organic carbon
• S - sulfur (S)
• zn - Zinc (Zn)
• fe - Iron (Fe)
• cu - Copper (Cu)
• Mn - Manganese (Mn)
• B - Boron (B)
• fertility: categorical (0 "Less Fertile", 1 "Fertile", 2 "Highly Fertile")

Resource/Situational Constraints

• Lack of Data about our region
• Lack of some classes
• Lack of experience in agriculutral science

Process followed

• Searched for a dataset from another country
• Creating additional instances for the least appearing class
• monitored by an experienced specialist

Code

1. Import liberaries and modules
2. Naive approuch with RandomForestClassifier and raw data(accuracy = 88%)
3. Data Exploration for modified data
4. Choose a model
   • SupportVectorClassifier
   • RandomForestClassifier
   • GaussianNB
   • KNeighborsClassifier
   • DecisionTreeClassifier
5. GridSearch
6. Train a RandomForestClassifier (accuracy = 97%)
7. Save the model