Merge branch 'main' of https://gitlab.stud.idi.ntnu.no/chrisrik/recommender-system into main

myfunc.py

+import matplotlib.animation as ani
+import matplotlib.cm as cmx
+import matplotlib.colors as colors
+import matplotlib.pyplot as plt
+import numpy as np
+import json
+import os
+import pandas as pd
+import statistics
+
+from sklearn.metrics.pairwise import cosine_similarity
+from matplotlib.patches import Ellipse
+from PIL import Image
+from sklearn import datasets
+from sklearn.cluster import KMeans
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+def load_data(path):
+    """
+        Load events from files and convert to dataframe.
+    """
+    map_lst=[]
+    for f in os.listdir(path):
+        file_name=os.path.join(path,f)
+        if os.path.isfile(file_name):
+            for line in open(file_name):
+                obj = json.loads(line.strip())
+                if not obj is None:
+                    map_lst.append(obj)
+    return pd.DataFrame(map_lst) 
+
+def load_dataset(df):
+    """
+        Convert dataframe to user-item-interaction matrix, which is used for 
+        Matrix Factorization based recommendation.
+        In rating matrix, clicked events are refered as 1 and others are refered as 0.
+    """
+    df = df[~df['documentId'].isnull()]
+    df = df.drop_duplicates(subset=['userId', 'documentId']).reset_index(drop=True)
+    df = df.sort_values(by=['userId', 'time'])
+    n_users = df['userId'].nunique()
+    n_items = df['documentId'].nunique()
+
+    ratings = np.zeros((n_users, n_items))
+    new_user = df['userId'].values[1:] != df['userId'].values[:-1]
+    new_user = np.r_[True, new_user]
+    df['uid'] = np.cumsum(new_user)
+    item_ids = df['documentId'].unique().tolist()
+    new_df = pd.DataFrame({'documentId':item_ids, 'tid':range(1,len(item_ids)+1)})
+    df = pd.merge(df, new_df, on='documentId', how='outer')
+    df_ext = df[['uid', 'tid']]
+    
+    for row in df_ext.itertuples():
+        ratings[row[1]-1, row[2]-1] = 1.0
+    return ratings
+
+def get_corr_list(user, M):
+    corr_list = []
+    for u in M.columns:
+        if u == user:
+            continue
+        A=M[user].to_numpy()
+        B=M[u].to_numpy()
+        result=cosine_similarity(A.reshape(1,-1),B.reshape(1,-1))
+        cor = result[0][0]
+        corr_list.append((u, cor))
+    corr_list.sort(key=lambda tup: tup[1], reverse=True)
+    return corr_list
+
+
+def get_rec(user, train_df, test_df, num, simular_users_count, filter=True):
+    users_list = [x[0] for x in user[1][:simular_users_count]]
+    simular_clicks_df = train_df[train_df['userId'].isin(users_list)]
+    user_clicks = test_df[test_df['userId'].isin([user])]['documentId'].to_list()
+    if filter:
+        df_unread_only = simular_clicks_df[~simular_clicks_df['documentId'].isin(user_clicks)]
+        titles = df_unread_only['documentId'].value_counts()
+    else:
+        titles = simular_clicks_df['documentId'].value_counts()
+    return titles[:num]
+
+
+
+def get_MSE(reality, prediction, articles):
+    # Given values
+    Y_true = np.zeros(len(articles))  # Y_true = Y (original values)
+    
+    # calculated values
+    Y_pred = np.zeros(len(articles))  # Y_pred = Y'
+
+    for a in reality:
+        Y_true[np.where(articles == a)] = 1
+
+    for a in prediction:
+        Y_pred[np.where(articles == a)] = 1
+    
+    # Calculation of Mean Squared Error (MSE)
+    r = mean_squared_error(Y_true,Y_pred)
+    return r
+
+def get_MSE2(reality, prediction, articles):
+    # Given values
+    Y_true = np.zeros(len(articles))  # Y_true = Y (original values)
+    
+    # calculated values
+    Y_pred = np.zeros(len(articles))  # Y_pred = Y'
+
+    for a in reality:
+        Y_true[np.where(articles == a)] = 1
+
+    for a in prediction:
+        Y_pred[np.where(articles == a)] = 1
+
+    new_y_true = []
+    new_y_pred = []
+    
+    for i in range(0, len(Y_true)):
+        if Y_true[i] == 1 or Y_pred[i] == 1:
+            new_y_true.append(Y_true[i])
+            new_y_pred.append(Y_pred[i])
+    
+    # Calculation of Mean Squared Error (MSE)
+    r = mean_squared_error(new_y_true,new_y_pred)
+    return r
\ No newline at end of file