Chemometrics and its Application in Pharmaceutical Field

Machine Learning for Process Monitoring and Control of Hot-Melt Extrusion: Current State of the Art and Future Directions

“…PCA is a technique for dimensionality reduction, which falls under unsupervised machine learning [ 65 , 76 , 77 ]. Details on how PCA works with PAT tools for monitoring pharmaceutical processes (other than HME) can be found in [ 78 , 79 , 80 ], and the detail of the algorithm is not repeated here. PCA has mostly been utilised in the HME literature to monitor the effect of varying processing conditions on the solid state of the drug.…”

“…Analogous to PCA, it involves a linear transformation of the data set, allowing for dimensionality reduction to a reduced number of ‘latent variables’ (LV), which are linear combinations of the original variables. General details on the workings of the PLS algorithm applied to PAT data for pharmaceutical process monitoring can be found in [ 78 , 79 , 80 , 83 ]. In pharmaceutical processes, PLS is primarily used to predict the concentration of the drug, although it has also been used to predict polymer blend contents, degradation of the polymer, the particle size of fillers in the polymer matrix, and mechanical properties of the polymer extrudate in non-pharma HME processes.…”