Apache Mahout Source Reading notes-datamodel Userbaserecommender

Let's take a look at the process of using: 1) Get DataModel2) define similarity calculation modelpearsoncorrelationsimilarity3) Define user neighborhood calculation modelNearestnuserneighborhood4) define the recommendation modelGenericuserbasedrecommender5) Make recommendations

@Test Public voidTesthowmany ()throwsException {datamodel Datamodel=Getdatamodel (New Long[] {1, 2, 3, 4, 5},            Newdouble[][] {{0.1, 0.2},                    {0.2, 0.3, 0.3, 0.6},                    {0.4, 0.4, 0.5, 0.9},                    {0.1, 0.4, 0.5, 0.8, 0.9, 1.0},                    {0.2, 0.3, 0.6, 0.7, 0.1, 0.2},            }); //for computing the most similar users, domain usersUsersimilarity similarity =Newpearsoncorrelationsimilarity (Datamodel); Userneighborhood Neighborhood=NewNearestnuserneighborhood (2, Similarity, Datamodel); Recommender Recommender=NewGenericuserbasedrecommender (Datamodel, neighborhood, similarity); List<RecommendedItem> fewrecommended = Recommender.recommend (1, 2); List<RecommendedItem> morerecommended = recommender.recommend (1, 4);  for(inti = 0; I < fewrecommended.size (); i++) {assertequals (Fewrecommended.get (i). Getitemid (), Morerecommended.get (i). Getitemid ()); } Recommender.refresh (NULL);  for(inti = 0; I < fewrecommended.size (); i++) {assertequals (Fewrecommended.get (i). Getitemid (), Morerecommended.get (i). Getitemid ()); }  }

Similarity calculation, refer to the pearsoncorrelationsimilarity of the previous article.

Nearestnuserneighborhood, how to get the nearest n users, how to achieve it?
~/mahout-core/src/main/java/org/apache/mahout/cf/taste/impl/recommender/genericuserbasedrecommender.java

@Override PublicList<recommendeditem> recommend (LongUseridintHowmany, Idrescorer Rescorer)throwstasteexception {preconditions.checkargument (Howmany>= 1, "Howmany must is at least 1"); Log.debug ("Recommending items for user ID ' {} '", UserID); //Calculates the most similar n users according to the similarity model    Long[] Theneighborhood =Neighborhood.getuserneighborhood (UserID); if(Theneighborhood.length = = 0) {      returncollections.emptylist (); }    //get a list of the item that is scored by users in other areas and not rated by the current user as the recommended base poolFastidset Allitemids =Getallotheritems (Theneighborhood, UserID); //get inside the pool and recommend the TOPN with the highest current user preferencesTopitems.estimator<long> Estimator =NewEstimator (UserID, Theneighborhood); List<RecommendedItem> Topitems =topitems. Gettopitems (Howmany, Allitemids.iterator (), rescorer, estimator); Log.debug ("Recommendations is: {}", Topitems); returnTopitems; }

The implementation of the estimator is this:

  Private Final classEstimatorImplementsTopitems.estimator<long> {        Private Final LongTheuserid; Private Final Long[] theneighborhood; Estimator (LongTheuserid,Long[] theneighborhood) {       This. Theuserid =Theuserid;  This. Theneighborhood =Theneighborhood; } @Override Public DoubleEstimate (Long ItemID)throwstasteexception {returndoestimatepreference (Theuserid, Theneighborhood, ItemID); }  }}

  protected floatDoestimatepreference (LongTheuserid,Long[] Theneighborhood,LongItemID)throwstasteexception {//add a similar user's preference to the item, and then average it as the current user's preference for changing the item    if(Theneighborhood.length = = 0) {      returnFloat.nan; } Datamodel Datamodel=Getdatamodel (); DoublePreference = 0.0; Doubletotalsimilarity = 0.0; intCount = 0;  for(LongUserid:theneighborhood) {      if(UserID! =Theuserid) {        //See Genericitembasedrecommender.doestimatepreference () tooFloat pref =Datamodel.getpreferencevalue (UserID, ItemID); if(Pref! =NULL) {          DoubleThesimilarity =similarity.usersimilarity (Theuserid, UserID); if(!Double.isnan (thesimilarity)) {Preference+ = Thesimilarity *pref; Totalsimilarity+=thesimilarity; Count++; }        }      }    }    //Throw out the estimate if it is based on no data points, of course, but also if based on//just one.    This was a bit of a Band-Aid on the ' stock ' item-based algorithm for the moment. //The reason is, and the estimate is, simply, the user's rating for one item//That happened to has a defined similarity. The similarity score doesn ' t matter, and that//seems a bad situation.    if(Count <= 1) {      returnFloat.nan; }    floatEstimate = (float) (Preference/totalsimilarity); if(Capper! =NULL) {estimate=capper.capestimate (estimate); }    returnestimate; }

Summarize:
1) Calculate the most similar n users
2) from the most similar n users, get the item that you have not scored
3) Anticipate your preference for each item
4) recommend the highest preference N item

Apache Mahout Source Reading notes-datamodel Userbaserecommender