Algorithm 310: Prime number generator 1

(l(~tts were .~ectio~ted according to bot,h the no~> prefere~ce a~d preference algorithms, attd the Immber of times they received their preferred sections i** courses with m~ltjt)le sections was tMlied. The results show that whet~ the >o**prefer(!,~ee algorithm is used a student receives at)o~*t 40 percent of lhe sections for which he has a prefer-e**ce, arid whe~, the preference algorithm is used this per-. ce~tage i~tereases to :wound 72 pereet~t, Both algori;hms meet the origit,al two requirements of m:~chit,e seeCio~ting equally well. A. subjective analysis of ihe resulting bahmee of sections when the preference algorithm was used shows it to be almost as good as the non-preference algorithm. Rarely was the imbalance more than oae student. 5. Conclusions Bec~mse of the small mtmber of courses where second preferences were permitted, the consequences of alternate courses on the Mgorithm cannot be fully determined. The oaly evidence that substar,tiates its effeetiveness is that t)E sections were balanced as well as sections ill non-PE courses. Further testing will probably show that ordering according to the number of seats left is not sufficient but ~hat the ordering should be done aeeording to the ratio of ~he number of seats left to seats originally available. Because of the relative unimportanee of second preferenees i, the data tested, ordering by ratio was not tried. A larger pereetttage of requests will probably decrease performance achieved if most requests are centered on only ~ few sections in eaeh course, sinee section sizes are limited. in the test data used this seemed to be the ease for some courses, l:'opular time periods may also cause trouble. For iastanee, there were only 811 requests for MWF at 8 ~/etoek classes while there were 1,390 requests for MWF at 9 o'clock classes. These numbers are only indicative of ~}:~e trend, since the effects of the time schedule on requests is itot readily analyzed. From the test ease tried, it appears that it is possible to allow students to have section preference with machine :~eetioning. The results achieved show that a student re-~;'eives around 32 pereent more of his preferred sections using the preference Mgorithnt-yet the prineiple objectives of machine registration are still satisfied. for their assistance in modifying and using the Washington State University maehiae registration program and for helpful discussions during development of the preference algorithm. integer procedure sievel@l, p); value m; integer m; i~l-teger array p; comment sievel(m, p) generates …