The feasibility and safety of a mining project or the choice among alternat
ive mining methods could depend on the joint densities and orientations wit
hin the rock mass. The accurate determination of the orientation of all joi
nts is technically difficult and often economically unrealistic. This study
presents a new approach in classifying joints found in exploration borehol
es as joint sets, whose statistical distribution is determined from a few h
undred oriented joints in boreholes. Each non-oriented joint is classified
as belonging to a set based on its "a posteriori" probability of membership
in a Bayesian framework. The theoretical rate of success of the classifica
tion can be computed for each possible borehole orientation and plotted on
a stereonet to determine the optimal orientation of new boreholes. The perf
ormance and limitations of this approach are investigated. An application e
xample at the Mont Porphyre's large scale block-caving project at Gaspe Min
es, Quebec, Canada, is studied.