Hybrid ensemble-based machine learning model for predicting phosphorus concentrations in hydroponic solution.

Area of Science:

• Agricultural Science
• Data Science
• Environmental Science

Background:

• Accurate phosphorus measurement is crucial for hydroponics to optimize plant growth and prevent environmental issues.
• Existing methods for phosphorus detection can be time-consuming or require labels, limiting their application.

Purpose of the Study:

• To evaluate the effectiveness of hybrid machine learning models for label-free phosphorus concentration prediction.
• To compare the performance of hybrid models against single machine learning models using absorbance data.

Main Methods:

• Employed three base classifiers: Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN).
• Utilized ensemble techniques (voting, bagging, stacking) to create hybrid models.
• Analyzed accuracy and computational time for both single and hybrid models.

Main Results:

• Support Vector Machine (SVM) achieved the highest accuracy (99.6%) among single models.
• The stacking hybrid model combining SVM, KNN, and RF yielded the highest accuracy (99.73%) with efficient computational time (36.18 s).
• Hybrid models demonstrated superior accuracy in predicting phosphorus levels compared to individual models.

Conclusions:

• Machine learning effectively distinguishes phosphorus concentrations in hydroponic systems.
• Hybrid machine learning techniques enhance prediction accuracy for phosphorus levels without requiring labels.
• The developed hybrid models offer a promising solution for rapid, label-free phosphorus monitoring in hydroponics.