def slicingUIM(U, I):
#Storing UIM values for indices in I and U for respective axes
M = UIM.iloc[I, U]
#Square and sum across users
M2 = (M * M).sum(0)
#Sort M and return the top n indices
return pd.argsort(M2)[:top_n]
def wildcard(uid):
mainList = pd.argsort(finalscore(uid))[:9].tolist()
dropList = []
dropList = dropList.append(mainList).append(genI(uid))
wildIIDList = pd.DataFrame(np.arange(5000)).drop(dropList)
upList = wildIIDList.apply(lambda i: len(genI(i)) + 1)
hm = scipy.stats.hmean(upList)
length = wildIIDList.size()
k = hm / length
pList = wildIIDList.apply(lambda i: k / (len(genI(i)) + 1))
wildIID = np.random.choice(wildIIDList, 1, True, pList)[0]
return random.shuffle(mainList.append(wildIID))
def icfscore(uid, iid):
I = initI_noIID(uid)
U = genU(iid)
#Storing top user indices
U1 = pd.argsort(UUM[uid, U])[:top_n].tolist()
return slicingUIM(U1, I)
def cbscore(uid, iid):
I = initI_noIID(uid)
#Storing the IIM values of this iid column restricted to indices in I
M = IIM.iloc[I, iid]
#Sort M and return the top n indices
return pd.argsort(M)[:top_n]