Open source software for numerical computation

Skip to main content

FACT - Free Access Chemometrics Toolbox

Jean-Claude Boulet, INRA

Chemometrics is widely developed in many industrial fields such as chemistry, drugs, food. The main applications consist in quantifying compounds using spectra, for example protein contents in wheat flour. Partial least square regression (PLS) is its most famous tool and has played an essential role in the development of chemometrics.

Nowadays most of the chemometric softwares are commercial, e.g. Unscrambler, PLS_toolbox, Pirouette. They offer the classical tools, a friendly interface and an efficient support for users, and that is valuable for industrial applications. On the other hand, they are not versatile and they do not contain the most recent developments or more confidential tools.

FACT - Free Access Chemometrics Toolbox allows to perform regressions with highly colinear variables, spectra pretreatments, discriminant and multi-block analysis. It is still updated with new functions. It was designed with the following properties:

under the Scilab platform: free, more advanced than Octave, equivalent to Matlab for the computations.
an expertise in the choice of the well-known methods: Methods are selected according to the bibliography; e.g. the PLS regression is represented by two algorithms over the nine that have been published.
a showcase of our own methods : Within a consortium, 6 French research teams and 2 private compagnies contribute to Fact. It is thus a support to communicate results from this consortium. It is also a platform for allowing comparisons of methods proposed by researchers.
a different audience: We wish to interest potential users that would not afford a commercial licence. In this way, students would freely access tools necessary to apply by themselves the methods they are learning. Researchers of public or private companies would get a versatile tool, suited to very different situations and with the possibility to be modified if necessary.
no technical support but enough help to get started : Fact help files are a reminder of the purpose of each function and its syntax. More information can be downloaded from Scilab : a tutorial about the use of Fact and its main functions, and the datasets used in the tutorial.
some guarantees of quality : Comparisons of results obtained from Fact vs commercial softwares will be also available soon.


Current contributors: J.C. Boulet and G. Mazerolles, INRA Montpellier; D. Bertrand, Data_Frame Nantes; J.M. Roger IRSTEA Montpellier.