K = 10
data = Dataset.load_builtin('ml-100k')
trainset = data.build_full_trainset()
trash, testset = train_test_split(data, test_size=.2)
# sim_options = {'name': 'pearson_baseline', 'user_based': False}
sim_options = {'user_based': False}
algo = KNNBaseline(sim_options=sim_options)
# algo = KNNBaseline()
D = 5 - 1
s = (trainset.n_users + 1, trainset.n_items + 1)
W = np.ones(s)
algo.weightUpdate(W)
predictions = algo.fit(trainset).test(testset)
# predictions = algo.fit(trainset)
PredictM = np.zeros(s)
for it in predictions:
PredictM[int(it[0]), int(it[1])] = it[3]
print(PredictM)
PM = pd.DataFrame(PredictM)
PM.to_csv("PredictionMatrix.csv")
# print (trainset.all_items())
NeighborM = np.zeros([trainset.n_items, K])
for i, item in enumerate(trainset.all_items()):
NeighborM[i] = algo.get_neighbors(item, k=K)
NM = pd.DataFrame(NeighborM)
# PM = pd.DataFrame(testset)
# PM.to_csv("TestSet.csv")
#
# # Use the famous SVD algorithm.
sim_options = {'name': 'pearson_baseline', 'user_based': False}
algo = KNNBaseline(sim_options=sim_options)
s = (sizefinder.n_users + 1, sizefinder.n_items + 1)
# trainM = np.zeros(s)
# for uid,iid,rating in sizefinder.all_ratings():
# trainM[uid][iid] = rating
# PM = pd.DataFrame(trainM)
# PM.to_csv("TrainSet.csv")
#
weight = np.ones(s)
algo.weightUpdate(weight)
# algo.fit(trainset)
# est,details,k_neighbor = [algo.estimate(uid,iid) for (uid, iid, _) in testset[1]]
# NeighborM = (trainset.n_users+1)*[(trainset.n_items+1)*[None]]
# for uid,iid,_ in trainset.all_ratings():
# _,_,K_neighbor = algo.estimate(uid,iid)
# NeighborM[uid][iid] = np.array(K_neighbor)[:,0][:10]
# uid = trainset.to_inner_uid(testset[1][0])
# iid = trainset.to_inner_iid(testset[1][1])
# print(len(trainset.ur[uid]))
# print(NeighborM[uid][iid])
# est = algo.estimate(uid,iid)
# print(est,isinstance(est, tuple))
