v-measure

tour_model_eval/confirmed_trips_eval_bins_clusters.py

@@ -9,65 +9,252 @@
 import emission.storage.decorations.analysis_timeseries_queries as esda
 import pandas as pd
 from numpy import *
+from sklearn import metrics
+from pandas.testing import assert_frame_equal
 
-# Spanish words to English
-span_eng_dict = {'revisado_bike':'test ride with bike','placas_de carro':'car plates','aseguranza':'insurance',
- 'iglesia':'church','curso':'course','mi_hija recién aliviada':'my daughter just had a new baby',
- 'servicio_comunitario':'community service','pago_de aseguranza':'insurance payment',
- 'grupo_comunitario':'community group','caminata_comunitaria':'community walk'}
-
-# Convert purpose
-map_pur_dict = {'course':'school','work_- lunch break':'lunch_break','on_the way home':'home',
-               'insurance_payment':'insurance'}
-
-# precision_bins takes five parameters
-# - all_bins_preci: the list that collects precision of each bin, should pass in an empty list
-# - sp2en=None means no need to translate language
-#   sp2en='True' will use span_eng_dict to change Spanish to English
-#
-# - cvt_purpose=None means no need to convert purposes
-#   cvt_purpose='True' will use map_pur_dict to convert purposes
-#   using this parameter should also set sp2en='True'
-def precision_bins (all_bins_preci,bins,non_empty_trips,sp2en=None,cvt_purpose=None):
-    for bin in bins:
-        bin_user_input = (non_empty_trips[i].data["user_input"] for i in bin if
-                          non_empty_trips[i].data["user_input"] != {})
-        bin_df = pd.DataFrame(data=bin_user_input)
-        if sp2en == 'True':
-            bin_df = bin_df.replace(span_eng_dict)
-        if cvt_purpose == 'True':
-            bin_df = bin_df.replace(map_pur_dict)
-        duplic_trips = bin_df[bin_df.duplicated(keep=False)]
-
-        # for bin that doesn't have duplicate trips, assign 0 as precision
-        if duplic_trips.empty and len(bin_df) > 1:
-            all_bins_preci.append(0)
-        # for bin only has one trip, assign 1.0 as precision
-        elif len(bin_df) == 1:
-            all_bins_preci.append(1.0)
+
+# - user_ls: a list of all users
+# - valid_user_ls: a list of valid users
+def get_user_ls(all_users,radius):
+    user_ls = []
+    valid_user_ls = []
+    for i in range(len(all_users)):
+        curr_user = 'user' + str(i + 1)
+        user = all_users[i]
+        filter_trips,sim,trips = filter_data(user,radius)
+        if valid_user(filter_trips,trips):
+            valid_user_ls.append(curr_user)
+            user_ls.append(curr_user)
         else:
-            duplic = duplic_trips.groupby(duplic_trips.columns.tolist()).apply(lambda x: tuple(x.index)).tolist()
-            max_duplic = max(duplic, key=lambda i: len(i))
-            precision = round(len(max_duplic) / len(bin), 2)
-            all_bins_preci.append(precision)
-    return all_bins_preci
-
-
-# precision_all_users takes four parameters
-# - all_users: pass in all participants' data
-# - sp2en: default None, no need to change language
-# - cvt_purpose: default None, no need to convert purpose
-def precision_bin_all_users(all_users,radius,sp2en=None,cvt_purpose=None):
-    all_users_preci = []
+            user_ls.append(curr_user)
+            continue
+    return user_ls,valid_user_ls
+
+
+# - trips: all trips read from database
+# - filter_trips: valid trips that have user labels and are not points
+def filter_data(user,radius):
+    trips = pipeline.read_data(uuid=user, key=esda.CONFIRMED_TRIP_KEY)
+    non_empty_trips = [t for t in trips if t["data"]["user_input"] != {}]
+    non_empty_trips_df = pd.DataFrame(t["data"]["user_input"] for t in non_empty_trips)
+    valid_trips_df = non_empty_trips_df.dropna(axis=0, how='any', thresh=None, subset=None, inplace=False)
+    valid_trips_idx_ls = valid_trips_df.index.tolist()
+    valid_trips = [non_empty_trips[i]for i in valid_trips_idx_ls]
+
+    # similarity codes can filter out trips that are points in valid_trips
+    sim = similarity.similarity(valid_trips, radius)
+    filter_trips = sim.data
+    return filter_trips,sim,trips
+
+
+# to determine if the user is valid:
+# valid user should have >= 10 trips for further analysis and the proportion of filter_trips is >=50%
+def valid_user(filter_trips,trips):
+    valid = False
+    if len(filter_trips) >= 10 and len(filter_trips) / len(trips) >= 0.5:
+        valid = True
+    return valid
+
+
+# to map the user labels
+# - user_input_df: pass in original user input dataframe, return changed user input dataframe
+# - sp2en: change Spanish to English
+# - cvt_pur_mo: convert purposes and replaced mode
+def map_labels(user_input_df,sp2en,cvt_pur_mo):
+    # Spanish words to English
+    span_eng_dict = {'revisado_bike': 'test ride with bike', 'placas_de carro': 'car plates', 'aseguranza': 'insurance',
+                     'iglesia': 'church', 'curso': 'course',
+                     'mi_hija recién aliviada': 'my daughter just had a new baby',
+                     'servicio_comunitario': 'community service', 'pago_de aseguranza': 'insurance payment',
+                     'grupo_comunitario': 'community group', 'caminata_comunitaria': 'community walk'}
+
+    # Convert purpose
+    map_pur_dict = {'course': 'school', 'work_- lunch break': 'lunch_break', 'on_the way home': 'home',
+                    'insurance_payment': 'insurance'}
+
+    if sp2en:
+        # change language
+        user_input_df = user_input_df.replace(span_eng_dict)
+    elif cvt_pur_mo:
+        # change language first
+        user_input_df = user_input_df.replace(span_eng_dict)
+        # convert purpose
+        user_input_df = user_input_df.replace(map_pur_dict)
+        # convert mode
+        for a in range(len(user_input_df)):
+            if user_input_df.iloc[a]["replaced_mode"] == "same_mode":
+                # to see which row will be converted
+                logging.debug("The following rows will be changed: %s", user_input_df.iloc[a])
+                user_input_df.iloc[a]["replaced_mode"] = user_input_df.iloc[a]['mode_confirm']
+    return user_input_df
+
+
+# check if the user is valid
+# append NaN to the score lists when the user invalid
+def valid_user_check(filter_trips,trips,homo_score,comp_score,v_score):
+    if not valid_user(filter_trips, trips):
+        homo_score.append(NaN)
+        comp_score.append(NaN)
+        v_score.append(NaN)
+        skip = True
+    else:
+        skip = False
+    return homo_score,comp_score,v_score,skip
+
+
+#  This function is to get homogeneity score, complete score, and v-score
+def compute_score(labels_true,labels_pred,homo_score,comp_score,v_score):
+    homo = metrics.homogeneity_score(labels_true, labels_pred)
+    homo_score.append(float('%.3f' % homo))
+    comp = metrics.completeness_score(labels_true, labels_pred)
+    comp_score.append(float('%.3f' % comp))
+    v = metrics.v_measure_score(labels_true, labels_pred)
+    v_score.append(float('%.3f' % v))
+    return homo_score,comp_score,v_score
+
+
+# v_measure_bins takes 5 parameters
+# - sp2en=True: change Spanish to English
+# - cvt_pur_mo=True: convert purposes and replaced mode
+# - cutoff=True: choose to analyze bins above cutoff
+# - cutoff=None: analyze all bins
+# Note: for sp2en and cvt_pur_mo, set either one to be True as needed. cvt_pur_mo will change language first
+def v_measure_bins(all_users,radius,sp2en=None,cvt_pur_mo=None,cutoff=None):
+    homo_score = []
+    comp_score = []
+    v_score = []
+    for i in range(len(all_users)):
+        user = all_users[i]
+        filter_trips,sim,trips = filter_data(user,radius)
+
+        homo_score,comp_score,v_score,skip = valid_user_check(filter_trips,trips,homo_score,comp_score,v_score)
+        if skip:
+            continue
+
+        sim.bin_data()
+        if cutoff is None:
+            trip_index_ls = []
+            bins = sim.bins
+            for bin in bins:
+                for index in bin:
+                    trip_index_ls.append(index)
+            bin_trips = [filter_trips[num] for num in trip_index_ls]
+
+        elif cutoff:
+            sim.delete_bins()
+            bin_trips = sim.newdata
+            bins = sim.bins
+
+        bin_trips_user_input_df = pd.DataFrame(data=[trip["data"]["user_input"] for trip in bin_trips])
+        bin_trips_user_input_df = map_labels(bin_trips_user_input_df, sp2en, cvt_pur_mo)
+
+        # turn all user_input into list without binning
+        bin_trips_user_input_ls = bin_trips_user_input_df.values.tolist()
+        # drop duplicate user_input
+        no_dup_df = bin_trips_user_input_df.drop_duplicates()
+        # turn non-duplicate user_input into list
+        no_dup_list = no_dup_df.values.tolist()
+
+        # collect labels_true based on user_input
+        labels_true = []
+        for trip in bin_trips_user_input_ls:
+            if trip in no_dup_list:
+                labels_true.append(no_dup_list.index(trip))
+
+        # collect labels_pred based on bins
+        labels_pred = []
+        for b in range(len(bins)):
+            for trip in bins[b]:
+                labels_pred.append(b)
+
+        # compare the trips order in bins and those in valid_trips using timestamp
+        bin_trips_ts = pd.DataFrame(data=[trip["data"]["start_ts"] for trip in bin_trips])
+        bin_ls = []
+        for bin in bins:
+            for index in bin:
+                bin_ls.append(index)
+        bins_ts = pd.DataFrame(data=[filter_trips[i]["data"]["start_ts"] for i in bin_ls])
+        # compare two data frames, the program will continue to score calculation if two data frames are the same
+        assert_frame_equal(bins_ts, bin_trips_ts)
+        homo_score, comp_score, v_score = compute_score(labels_true, labels_pred, homo_score, comp_score, v_score)
+
+    return homo_score, comp_score, v_score
+
+
+# - sp2en=True: change Spanish to English
+# - cvt_pur_mo=True: convert purposes and replaced mode
+# - cutoff=True: choose to analyze bins above cutoff
+# - cutoff=None: analyze all bins
+# Note: for sp2en and cvt_pur_mo, set either one to be True as needed. cvt_pur_mo will change language first
+def v_measure_clusters(all_users,radius,sp2en=None,cvt_pur_mo=None):
+    homo_score = []
+    comp_score = []
+    v_score = []
     for i in range(len(all_users)):
         user = all_users[i]
-        trips = pipeline.read_data(uuid=user, key=esda.CONFIRMED_TRIP_KEY)
-        all_bins_preci = []
-        non_empty_trips = [t for t in trips if t["data"]["user_input"] != {}]
-        if non_empty_trips != {}:
-            sim = similarity.similarity(non_empty_trips, radius)
-            if sim.data:
-                sim.bin_data()
-                all_bins_preci = precision_bins(all_bins_preci, sim.bins, non_empty_trips, sp2en, cvt_purpose)
-        all_users_preci.append(round(mean(all_bins_preci), 2))
-    return all_users_preci
+        filter_trips,sim,trips = filter_data(user,radius)
+
+        homo_score,comp_score,v_score,skip = valid_user_check(filter_trips,trips,homo_score,comp_score,v_score)
+        if skip:
+            continue
+
+        sim.bin_data()
+        sim.delete_bins()
+        bin_trips = sim.newdata
+        bins = sim.bins
+
+        # clustering the data only based on sil score (min_cluster = 0) instead of bins number (len(bins))
+        feat = featurization.featurization(bin_trips)
+        min = 0
+        max = int(math.ceil(1.5 * len(bins)))
+        feat.cluster(min_clusters=min, max_clusters=max)
+        cluster_trips = feat.data
+        cluster_user_input_df = pd.DataFrame(data=[i["data"]["user_input"] for i in cluster_trips])
+        cluster_user_input_df = map_labels(cluster_user_input_df, sp2en, cvt_pur_mo)
+        # turn cluster_trips to list without any changes
+        cluster_user_input_ls = cluster_user_input_df.values.tolist()
+        # drop duplicate user_input
+        no_dup_df = cluster_user_input_df.drop_duplicates()
+        # turn non-duplicate user_input into list
+        no_dup_list = no_dup_df.values.tolist()
+        # collect labels_true based on user_input
+        labels_true = []
+        for trip in cluster_user_input_ls:
+            if trip in no_dup_list:
+                labels_true.append(no_dup_list.index(trip))
+        labels_pred = feat.labels
+
+        # compare the points in cluster_trips and those in feat.points, the program will continue to score calculation
+        # if the frames are the same
+        cluster_ps = []
+        for trip in cluster_trips:
+            cluster_ps.append([trip["data"]["start_loc"]["coordinates"][0],
+                               trip["data"]["start_loc"]["coordinates"][1],
+                               trip["data"]["end_loc"]["coordinates"][0],
+                               trip["data"]["end_loc"]["coordinates"][1]])
+        cluster_ps_df = pd.DataFrame(data=cluster_ps)
+        label_ps_df = pd.DataFrame(data=feat.points)
+        assert_frame_equal(cluster_ps_df, label_ps_df)
+        homo_score, comp_score, v_score = compute_score(labels_true, labels_pred, homo_score, comp_score, v_score)
+
+    return homo_score, comp_score, v_score
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+