Page 71 - Veritas Vol 3, Issue 2
P. 71

61





             conscientious procedures for handling, storing and analysing, there requires a

             more  reliable  approach.  The  continuous  efforts  made  in  recent  decades  in
             establishing  a  widely  accepted  approach  has  resulted  in  this  fruitful  method

             called  Chemometrics  which  involves  statistical  inference,  allowing  more
             improved probabilistic interpretation. It can help visualise complex analytical
             data in a simpler and more convenient manner. As we evolve into the modern

             world, the amount of data that is generated everywhere is vast. Almost every
             process is computerised and visualised, and there is necessity for the forensic

             scientist to understand these complex tools for viewing the collected data more
             seamlessly.  In  simple  words  chemometrics  is  the  integration  of  statistics,
             mathematics  and  computational  methods  to  extract  relevant  chemical

             information from complex data obtained during investigation.



             2. CHEMOMETRIC METHODS



             Techniques in chemometrics varies based on the application on which it is being
             used.  There  are  three  different  methods,  Pattern  Recognition,  which  is

             subdivided  into  Unsupervised  and  Supervised  method,  Regression  Technique
             and Experimental Design met



             2.1 Pattern Recognition:
             2.1.1 Unsupervised Pattern Recognition

             This method mainly involves detecting patterns in the datasets of the collected
             samples without any predefined labels or outcomes, with the help of algorithms

             which  largely  decreases  human  error.  This  method  also  allows  comparative
             analysis  with  a  process  of  projecting  the  new  samples  onto  the  previously

             collected samples. Hierarchical Cluster Analysis (HCA) which involves merging
             data  into  large  clusters  based  on  special  distance  and  relative  similarities,
             Principal  Component  Analysis  (PCA)  which  uses  dimensionality  reduction

             technique  to  offer  visual  representation  of  the  data  are  examples  of
             Unsupervised Pattern Recognition technique.

             2.1.2 Supervised Pattern Recognition
             In a supervised method, data used involves labelled data. As the name suggests

             it involves human intervention which makes it more reliable and acceptable in
             the  court  of  law  as  there  is  a  proof  that  the  process  has  been  done  under

             authorised supervision.




     VERITAS VOLUME: 3, ISSUE: 2                                            WRITER: R V ISHITHA REDDY
   66   67   68   69   70   71   72   73   74   75   76