Nitrate Classification Based on Optical Absorbance Data Using Machine Learning Algorithms for a Hydroponics System.

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Environmental Science

Background:

• Accurate nitrate (NO3-) determination in hydroponic nutrient solutions is critical for optimizing plant growth and preventing environmental pollution.
• Both deficient and excessive nitrate levels can negatively impact crop yield and water quality.

Purpose of the Study:

• To evaluate the effectiveness of feature reduction techniques (LDA, PCA) and machine learning (ML) algorithms for quantifying nitrate concentration.
• To compare the performance of PCA and LDA in conjunction with various ML models using high-dimensional spectroscopic data.

Main Methods:

• Utilized a high-dimensional spectroscopic dataset of nitrate-nitrite mixed solutions.
• Applied two feature reduction techniques: Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA).
• Evaluated seven machine learning algorithms: KNN, SVM, Decision Trees, Naive Bayes, Random Forest (RF), Gradient Boosting, and Extreme Gradient Boosting.

Main Results:

• Principal Component Analysis (PCA) significantly outperformed Linear Discriminant Analysis (LDA) for feature reduction on the spectroscopic dataset.
• Machine learning models integrated with PCA achieved classification accuracies ranging from 92.7% to 99.8%.
• The combination of PCA and the Random Forest (RF) algorithm demonstrated the highest accuracy at 99.8%.

Conclusions:

• Feature reduction, especially PCA, is highly effective for improving nitrate concentration analysis in hydroponics.
• Machine learning algorithms, particularly Random Forest, coupled with PCA, offer a robust and accurate method for nitrate detection in complex solutions.
• This approach holds promise for precision agriculture and environmental monitoring applications.