get vector

personas_vector/to_vector.py

@@ -3,38 +3,41 @@ import sys
 
 sys.path.append(os.path.realpath("."))
 
+import multiprocessing
+
 import pandas as pd
 from gensim.models import Word2Vec, word2vec
 from utils.defs import nth_element
 from utils.files import get_df
 
-DEVICE_COLUMNS = [
-    "device_id",
-]
 
-TRACTATE_COLUMNS = [
-    "card_id",
-]
+def device_tractate_fe():
+    device_tags_df = get_df("personas_device_feature.csv")
+    device_tags_df = device_tags_df[["cl_id", "business_tags"]]
+    device_tags_df["business_tags"] = device_tags_df["business_tags"].str.split(",").\
+        apply(lambda d: d if isinstance(d, list) else [])
+    print(device_tags_df.head(3))
 
+    tractate_tags_df = get_df("personas_tractate_tags.csv", columns=["tractate_id", "business_tags"])
+    tractate_tags_df["business_tags"] = tractate_tags_df["business_tags"].str.split(",").\
+        apply(lambda d: d if isinstance(d, list) else [])
+    print(tractate_tags_df.head(3))
 
-def device_tractae_fe():
-    pass
+    return device_tags_df, tractate_tags_df
 
 
-if __name__ == "__main__":
-    device_fe_df = get_df("personas_device_feature.csv",
-                          columns=[
-                              "cl_id",
-                              "first_demands",
-                              "first_solutions",
-                              "first_positions",
-                              "second_demands",
-                              "second_solutions",
-                              "second_positions",
-                              "projects",
-                              "business_tags",
-                          ])
-    print(device_fe_df.head(3))
+def tractate_business_tags_word2vec(tractate_df):
+    data = tractate_tags_df["business_tags"].to_list()
+    print(len(data))
+    model = Word2Vec(data, hs=0, min_count=3, workers=multiprocessing.cpu_count(), iter=10)
+    print(model)
+    return model
 
-    tractate_tags_df = get_df("personas_tractate_tags.csv", columns=["tractate_id", "business_tags"])
-    print(tractate_tags_df.head(3))
+
+if __name__ == "__main__":
+    device_tags_df, tractate_tags_df = device_tractate_fe()
+    model = tractate_business_tags_word2vec(tractate_tags_df)
+    for i in ["自体脂肪面部年轻化", "自体脂肪填充面部", "自体脂肪全面部填充", "自体脂肪面部填充", "鼻综合", "鼻部综合"]:
+        print(model.wv.most_similar(i))
+        print(model.wv.get_vector(i))
+    # tractate_tags_df["business_tags"].to_list()

utils/files.py

@@ -34,7 +34,11 @@ def save_dict_to_csv(d, file):
 
 def get_df(file, sep="|", columns=[]):
     full_path = os.path.join(DATA_PATH, file)
+    # full_path = os.path.join("/Users/offic/work/GM/strategy_embedding/_data", file) # TODO
     print(full_path)
-    df = pd.read_csv(full_path, sep="|", names=columns)
+    if columns:
+        df = pd.read_csv(full_path, sep=sep, names=columns)
+    else:
+        df = pd.read_csv(full_path, sep=sep)
     print(df.shape)
     return df