Machine Learning-Based Software for Predicting Pseudomonas spp. Growth Dynamics in Culture Media

Area of Science:

• Predictive Microbiology
• Food Safety
• Bacterial Growth Modeling

Background:

• Traditional primary and secondary models are used to estimate microbial growth.
• Pseudomonas spp. are key bacteria in food spoilage.
• Predictive modeling aids in food safety and shelf-life estimation.

Purpose of the Study:

• To develop a machine learning tool for predicting Pseudomonas spp. growth.
• To compare the performance of machine learning models against traditional growth models.
• To identify the most accurate machine learning approach for microbial growth prediction.

Main Methods:

• Applied Support Vector Regression (SVR), Random Forest Regression (RFR), and Gaussian Process Regression (GPR) models.
• Used temperature, water activity, and pH as predictor variables.
• Compared machine learning models with Gompertz, Logistic, Baranyi, and Huang models using R2adj and RMSE.

Main Results:

• Machine learning models demonstrated superior accuracy (R2adj: 0.834–0.959; RMSE: 0.005–0.010) compared to traditional methods.
• Gaussian Process Regression (GPR) was the most accurate model for both training and testing.
• External validation confirmed GPR's reliability (Bf: 0.998–1.047; Af: 1.100–1.167).

Conclusions:

• Machine learning models, particularly GPR, offer a more accurate approach to predicting microbial growth.
• This method bypasses the need for secondary modeling, simplifying predictive microbiology.
• The developed tool provides a robust alternative for microbial growth prediction in food safety applications.