Matlab 8.3.0.532 r2014a

Survival kinetic data of S. Reading on the fresh iceberg lettuce stored at 4 and 8 °C were modelled using Weibull model as described by Mafart et al. [26 (link)]: $N\left(t\right)=N_0-{\left(\frac{t}{delta}\right)}^p$
where t is the time (h), N₀ is the concentration of microbial population (log₁₀ CFU/g) at time t, delta is the scale parameter and p is shape parameter.
Similar to the Baranyi model, the parameters of the Weibull model were calculated with the NonLinearModel.fit command in the Matlab 8.3.0.532 (R2014a) software (MathWorks Inc., Natick, MA, USA).

The UV- Visible spectrophotometer JASCO dual beam (Tokyo, Japan) model V-630 was utilized with the included programmer spectra II manager. The spectral slit had a width of 2 nm, and the scanning rate was 1000 nm/min. All chemometric techniques were applied using MATLAB^®8.3.0.532 (R2014a), PLS Toolbox (version 2.1), MCR-ALS Toolbox, and Neural Network Toolbox (Math Works, United States).

The bacterial count of Pseudomonas spp. in the unit of log CFU was defined as the main objective function considering the entire dataset categorised into numerical and categorical values for each record ID. The parameters “time”, “temperature”, “NaCl concentration”, “water activity”, and “CO₂ concentration” are numerical data. The microbial counts (log CFU/g) at 0 h were determined as the initial count of Pseudomonas spp. for each record ID. To separate initial counts from others, data belonging to a time of 0 (h) were coded as 0, and other data were coded as 1. Through this process, the information on the initial count of Pseudomonas spp. was also converted to numerical data. The parameters, vacuum condition (yes/no), food category (beef, pork, poultry, and culture medium), and food name (minced beef, pork, raw meat lombo, turkey, brain heart infusion broth “BHIB”, and several kinds of tryptic soy broth “TSB”), are categorical data and were kept as is. These variables were not transformed into numerical values, and they were directly used for predictions to avoid the possibility that the machine learning algorithms can create bias in the encoded variables by assuming that higher numbers are more important. The pre-processing steps were performed using Matlab 8.3.0.532 (R2014a) software (MathWorks Inc., Natick, MA, USA).