Page 31 - Banking Finance December 2018

A

co ercia statistica
soft are s ch as
ie s Mat a etc. or
the free open so rce
soft are R . rther an
artificia ne ra s ste
hich si ates the
h an earnin process
i ht e e en ore
sef . he s ste
i ur t rn tiv t vis vis r ition t earns the nat re of the re ationship et een inp ts and
o tp ts repeated sa p in inp t o tp t infor ation
rther a ar e proportion of the ne a ai a e data is sets. e ra net or s ha e a partic ar ad anta e o er
passi e prod ced as a res t of peop e s interactions ith e pert s ste s hen data are nois or inco p ete these
di ita ser ices s ch as o i e phones internet searches can a e an ed cated ess as ch as o d a h an
on ine p rchases and e ectronic pa ent transactions. e pert.
haracteristics of indi id a s can e inferred de e opin
a orith s that a e se of these data. s the pri ar a e ohnson and Raina descri ed ho ne ra
enerators of the data are not in the contro of their di ita net or s can incorporate s ecti e non antifia e
footprints and a so cannot ide the a s in hich it a infor ation into credit appro a decisions. i and cott
e sed there is hard an ris of anip ated or sed a s per ised artificia ne ra net or to predict
ca o f a ed infor ation as in the case of the traditiona an r ptc in a sa p e of fir s. herefore the
financia state ents and pro ections. o e er there re ain cha en e is to choose a set of a ternati e data points that
concerns a o t the pri ac . hi e pri ac can e protected to ether create a co prehensi e ris profi e for the
s ita e safe ards the access to hi h a it orro er si i ar to the pop ar fi e s of credit character
nad terated data and a i it to a e decisions ased on capacit capita co atera and conditions. ch ode s are
s ch data is a reat acco p ish ent in the cha en in or d eas to nderstand and are consistent ith the standard ris
of credit ana tics. ana e ent fra e or s.

. sin ternati e ata an the . Re ent e e op ents in a ternati e

ssess ent pproa hes re it s orin
t is e pected that the a o nt of data in the or d i ar e n ers of o inco e peop e icro entreprene rs
increase ti es et een 2 and 2 2 . o e er a s a sca e sinesses and r ra pop ations that pre io s
a orit of this data is non inear and nstr ct red in the did not ha e access to for a financia ser ices are no
for of i a es doc ents and ideos. n a ternati e increasin ein di ita an ed a ran e of o d and ne
approach to credit ana sis therefore re ires data inin financia ser ices pro iders inc din non traditiona
and sin ad anced a orith s for findin re ationships in a pro iders s ch as o i e net or operators B s and
non inear incon r ent data set. tho h a p re achine intechs. o o in are so e of the s ccess stories fro
earnin approach i ht res t in i pro ed prediction in across the or d hich sho s the i ense potentia of the
these sit ations o er a h an earnin process there is a so a ternati e credit ana tics.
erit in the ar ent that the siness and ris ana ers st d cond cted researchers at the Massach setts
st nderstand as to ho the inferences are dra n. ross nstit te of echno o M has de onstrated that
ta ation or si i ar ana sis of sin e predictors is the core the introd ction of techno o ena ed financia
i din oc of the a ternati e credit scorin ode s. ser ices can he p red ce po ert . he st d esti ated
reatin cross ta ations is not er diffic t sin an that since 2 access to o i e one ser ices that

B CE | DECEMBER | 201 | 1