A simple Python-based referral system

defloadexdata ():return[[1,1,1, 0,0], [2,2,2, 0,0], [1,1,1, 0,0], [5,5,5, 0,0], [1,1,0,2,2], [0,0,0,3,3], [0,0,0,The]]defloadExData2 ():return[[0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 5], [0, 0, 0,3, 0, 4, 0, 0, 0, 0, 3], [0, 0, 0, 0,4, 0, 0, 1, 0, 4, 0], [3, 3, 4, 0, 0, 0, 0, 2, 2, 0, 0], [5, 4, 5, 0, 0, 0, 0, 5, 5, 0, 0], [0, 0, 0, 0,5, 0, 1, 0, 0, 5, 0], [4, 3, 4, 0, 0, 0, 0, 5, 5, 0, 1], [0, 0, 0,4, 0, 4, 0, 0, 0, 0, 4], [0, 0, 0,2, 0, 2, 5, 0, 0, 1, 2], [0, 0, 0, 0,5, 0, 0, 0, 0, 4, 0], [1, 0, 0, 0, 0, 0, 0, 1, 2, 0, 0]]  fromNumPyImport* fromNumPyImportLinalg as La#Euclidean distancedefEuclidsim (INA,INB):return1.0/(1.0+la.norm (ina-InB))#Pearson correlation coefficientdefPearssim (INA,INB):ifLen (InA) <3:return1.0return0.5+0.5*corrcoef (ina,inb,rowvar=0) [0][1]#Cosine similarity degreedefCossim (INA,INB): Num=float (ina.t*InB) Denom=la.norm (InA) *la.norm (InB)return0.5+0.5* (num/denom)#recommendation engine based on similarity of items (user estimates under the standard similarity calculation method)defstandest (datamat,user,simmeas,item):#Number of itemsN=shape (Datamat) [1]    #two variables to calculate the estimated scoring valuesimtotal=0.0;ratsimtotal=0.0#traverse all items and compare it to all items     forJinchrange (n):#user's rating of an itemuserrating=Datamat[user,j]ifUserrating==0:Continue        #Logical_and: The matrix is run logically with each element, and the return value is true,false for each element        #Datamat[:,item]. A>0: An element greater than 0 in the item column        #Datamat[:,j]. A: Elements greater than 0 in column J        #The row subscript (a vector) of the element that is both greater than 0 in Overlap:datamat[:,item],datamat[:,j]Overlap=nonzero (Logical_and (Datamat[:,item). A>0, Datamat[:,j]. A>0)) [0]Print(j)Print("------overlap------")        Print(overlap)ifLen (overlap) ==0:similarity=0#calculate the similarity of overlap matrices        Else: similarity=Simmeas (Datamat[overlap,item], datamat[overlap,j])Print("Datamat[overlap,item:")        Print(Datamat[overlap,item])Print("datamat[overlap,j:")        Print(Datamat[overlap,j])Print('The %d and%d similarity is:%f'%(item,j,similarity))#Cumulative total similarity (not quite understood)#Suppose A score is unknown, A, B similarity 0.9,b score 5,;a C similarity 0.8,c rating 4.#then according to the formula a score = (0.9*5+0.8*4)/(0.9+0.8)#equivalent to a weighted average (if divided by 2), but because the weights of the 2 ratings are different, the sum of the similarity should be divided bysimtotal+=Similarity#ratsimtotal = similarity * Element valueRatsimtotal+=similarity*userratingPrint("ratsimtotal+=similarity*userrating:")        Print(ratsimtotal)ifSimtotal==0:return0Else:returnratsimtotal/Simtotal#produce the highest n recommended results for a user#user indicates the number of users to recommenddefRecommend (datamat,user,n=3,simmeas=cossim,estmethod=standest):#Create a matrix of non-graded items for a given userUnrateditems=nonzero (Datamat[user,:]. a==0) [1]#element equal to 0 in the first user line#print (Datamat[user,:]. a==0)----[[True True True] ..., true False true]]#for a two-dimensional array B2,nonzero (B2), a tuple of length 2 is obtained. Its No. 0 element is the subscript of the No. 0 axis of an element with a value not 0 in array A, and the 1th element is the subscript of the 1th axis, so the values of b2[0,0], b[0,2], and b2[1,0] are not 0 from the following results:##>>> b2 = Np.array ([[True, False, True], [True, False, false]])#>>> Np.nonzero (B2)#(Array ([0, 0, 1], Dtype=int64), array ([0, 2, 0], Dtype=int64))       ifLen (unrateditems) ==0:return 'You rated everything'    #A list of projected scores for a non-rated itemitemscores=[]     forIteminchUnrateditems:#predict scores by Standest () method for each non-rated item        Print("Item------------")        Print(item) Estimatedscore=Estmethod (Datamat,user,simmeas,item)#Store item numbers and estimated scores in the listItemscores.append ((item,estimatedscore))#sorted sort function, key is sorted by keyword, lambda is implicit function, fixed notation,    #JJ means to sort the progenitor, jj[1] According to JJ's second column sort, reverse=true, descending; [: n] Top n    returnSorted (itemscores,key=LambdaJj:jj[1],reverse=True) [: N]#using SVD to improve the recommended effect#estimation of scoring based on SVDdefsvdest (datamat,user,simmeas,item):#Number of itemsN=shape (Datamat) [1] Simtotal=0.0;ratsimtotal=0.0#SVD decomposition is: u*s*vu,sigma,vt=LA.SVD (Datamat)#after decomposition, use only 90% of the singular value of energy, stored in the numpy arraySig4=mat (Eye (4) *sigma[:4])    #converting items into low-dimensional space using the U matrixxformeditems=datamat.t*u[:,:4]*sig4.i forJinchrange (N): userrating=Datamat[user,j]ifUserrating==0orJ==item:ContinueSimilarity=Simmeas (Xformeditems[item,:]. T, Xformeditems[j,:]. T)Print('The %d and%d similarity is:%f'%(item,j,similarity)) Simtotal+=Similarity Ratsimtotal+=similarity*userratingifSimtotal==0:return0Else:returnratsimtotal/Simtotalif __name__=='__main__': Mymat=Mat (LoadExData2 ())Print(Recommend (mymat,2))

　　

A simple Python-based referral system