from utils.datareader import Datareader import scipy.sparse as sps import sys from utils.definitions import ROOT_DIR arg = sys.argv[1:] mode = arg[0] dr = Datareader(verbose=False, mode=mode, only_load=True) top_pop = dr.get_eurm_top_pop(top_pop_k=750, remove_duplicates=True, binary=True) sps.save_npz(ROOT_DIR+"/recommenders/script/main/"+mode+"_npz/top_pop.npz", top_pop)
ucm = sparse.csr_matrix((data, (rows, cols)),
shape=(max(self.playlists) + 1,
len(list(self.title_to_idx.keys()))))
return ucm
if __name__ == '__main__':
datareader = Datareader(mode='offline', only_load=True)
evaluator = Evaluator(datareader)
nlp_lele = sparse.load_npz(ROOT_DIR +
'/data/ensemble/nlp_eurm_offline_bm25.npz')
nlp_strict = sparse.load_npz(ROOT_DIR + '/data/eurm_nlp_strict.npz')
top_pop = datareader.get_eurm_top_pop()
top_pop = norm_l1_row(top_pop)
nlp_lele = norm_l1_row(nlp_lele)
nlp_strict = norm_l1_row(nlp_strict)
nlp_fusion = (nlp_lele * 0.6) + (nlp_strict * 0.4)
sparse.save_npz(ROOT_DIR + '/data/eurm_nlp_fusion_offline.npz', nlp_fusion)
evaluator.evaluate(eurm_to_recommendation_list(nlp_fusion,
datareader=datareader),
name='nlp_fusion_no_toppop')
for a in [0.50, 0.55, 0.60, 0.65, 0.70]:
for b in [0.10, 0.15]:
nlp_fusion = (nlp_lele * a) + (nlp_strict *
(1.0 - a)) + (b * top_pop)
