Commit efb902224ab2245fac5ea7c46801af7e8d4e448a
1 parent
8ce5c499
Exists in
master
and in
1 other branch
Added weighting scheme and clustering options to recommender; changed method for…
… setting recommender strategy; do not use eval due to security risks; implemmented tests for recommender.
Showing
1 changed file
with
27 additions
and
42 deletions
Show diff stats
src/recommender.py
... | ... | @@ -19,10 +19,10 @@ __license__ = """ |
19 | 19 | along with this program. If not, see <http://www.gnu.org/licenses/>. |
20 | 20 | """ |
21 | 21 | |
22 | -from operator import itemgetter | |
23 | -from data import * | |
24 | -from strategy import * | |
25 | -from error import Error | |
22 | +import xapian | |
23 | +import operator | |
24 | +import data | |
25 | +import strategy | |
26 | 26 | |
27 | 27 | class RecommendationResult: |
28 | 28 | """ |
... | ... | @@ -40,7 +40,7 @@ class RecommendationResult: |
40 | 40 | """ |
41 | 41 | result = self.get_prediction() |
42 | 42 | str = "\n" |
43 | - for i in range(len(result)): | |
43 | + for i in range(len((list(result)))): | |
44 | 44 | str += "%2d: %s\n" % (i,result[i][0]) |
45 | 45 | return str |
46 | 46 | |
... | ... | @@ -48,8 +48,10 @@ class RecommendationResult: |
48 | 48 | """ |
49 | 49 | Return prediction based on recommendation size (number of items). |
50 | 50 | """ |
51 | - sorted_result = sorted(self.item_score.items(), key=itemgetter(1)) | |
52 | - return reversed(sorted_result[-size:]) | |
51 | + if size > len(self.item_score): size = len(self.item_score) | |
52 | + sorted_result = sorted(self.item_score.items(), | |
53 | + key=operator.itemgetter(1)) | |
54 | + return list(reversed(sorted_result[-size:])) | |
53 | 55 | |
54 | 56 | class Recommender: |
55 | 57 | """ |
... | ... | @@ -59,47 +61,30 @@ class Recommender: |
59 | 61 | """ |
60 | 62 | Set initial parameters. |
61 | 63 | """ |
62 | - try: | |
63 | - strategy = "self."+cfg.strategy+"(cfg)" | |
64 | - exec(strategy) | |
65 | - except (NameError, AttributeError, SyntaxError) as err: | |
66 | - print err | |
67 | - logging.critical("Could not perform recommendation strategy '%s'" % | |
68 | - cfg.strategy) | |
69 | - raise Error | |
70 | - | |
71 | - def ct(self,cfg): | |
72 | - """ | |
73 | - Set recommender attributes to perform content-based recommendation | |
74 | - using tags index as source data. | |
75 | - """ | |
76 | - self.items_repository = TagsXapianIndex(cfg) | |
77 | - self.strategy = ContentBasedStrategy() | |
78 | - | |
79 | - def cta(self,cfg): | |
80 | - """ | |
81 | - Set recommender attributes to perform content-based recommendation | |
82 | - using apt-xapian-index as source data. | |
83 | - """ | |
84 | 64 | self.items_repository = xapian.Database(cfg.axi) |
85 | - self.strategy = AxiContentBasedStrategy() | |
86 | - | |
87 | - def col(self,cfg): | |
88 | - """ | |
89 | - Set recommender attributes to perform collaborative recommendation | |
90 | - using popcon-xapian-index as source data. | |
91 | - """ | |
92 | - self.users_repository = PopconXapianIndex(cfg) | |
93 | - self.strategy = CollaborativeStrategy() | |
65 | + self.users_repository = data.PopconXapianIndex(cfg) #[FIXME] only cfg fields | |
66 | + self.clustered_users_repository = data.PopconXapianIndex(cfg) #[FIXME] | |
67 | + self.set_strategy(cfg.strategy) | |
68 | + if cfg.weight == "bm25": | |
69 | + self.weight = xapian.BM25Weight() | |
70 | + else: | |
71 | + self.weight = xapian.TradWeight() | |
94 | 72 | |
95 | - def set_strategy(self,strategy): | |
73 | + def set_strategy(self,strategy_str): | |
96 | 74 | """ |
97 | 75 | Set the recommendation strategy. |
98 | 76 | """ |
99 | - self.strategy = strategy | |
77 | + if strategy_str == "cb": | |
78 | + self.strategy = strategy.ContentBasedStrategy("full") | |
79 | + if strategy_str == "cbt": | |
80 | + self.strategy = strategy.ContentBasedStrategy("tag") | |
81 | + if strategy_str == "cbd": | |
82 | + self.strategy = strategy.ContentBasedStrategy("desc") | |
83 | + if strategy_str == "col": | |
84 | + self.strategy = strategy.CollaborativeStrategy(20) | |
100 | 85 | |
101 | - def get_recommendation(self,user): | |
86 | + def get_recommendation(self,user,limit=20): | |
102 | 87 | """ |
103 | 88 | Produces recommendation using previously loaded strategy. |
104 | 89 | """ |
105 | - return self.strategy.run(self,user) | |
90 | + return self.strategy.run(self,user,limit) | ... | ... |