Geometric search: A new approach for fitting PARAFAC2 models on GC-MS data

Area of Science:

• Chemometrics
• Analytical Chemistry
• Data Analysis

Background:

• PARAFAC2 is a decomposition technique well-suited for gas chromatography-mass spectrometry (GC-MS) data.
• Current fitting algorithms, like alternating least squares (ALS), are computationally slow, limiting PARAFAC2's practical application.
• Efficient modeling of complex GC-MS data is crucial for accurate sample analysis.

Purpose of the Study:

• To introduce and evaluate a novel iterative method, geometric search, for fitting the PARAFAC2 model.
• To enhance the convergence speed and fitting accuracy of PARAFAC2 decomposition.
• To address the computational limitations of existing PARAFAC2 fitting algorithms.

Main Methods:

• Developed an iterative geometric search algorithm to fit PARAFAC2 models.
• Modeled PARAFAC2 loading parameters as geometric sequences with offsets during ALS iterations.
• Evaluated the method using simulated datasets and real-world GC-MS data from wine and tobacco samples.

Main Results:

• The geometric search method demonstrated significantly faster convergence compared to standard ALS and line search algorithms.
• The proposed method achieved superior or comparable fitting quality to existing algorithms.
• Performance was validated on both simulated and complex real-world GC-MS datasets.

Conclusions:

• Geometric search provides an efficient and effective approach for fitting PARAFAC2 models.
• This method overcomes the speed limitations of traditional ALS, enabling broader use of PARAFAC2 for GC-MS data analysis.
• The enhanced fitting performance has implications for improved data interpretation in fields utilizing GC-MS.