Duplicates found in y_true list

Current Version : 0.4.0

In function print_metrics_ranking, I have added in the lines:

        rec_list = pd.DataFrame.from_dict(y_reco_list, orient="index").reset_index()
        rec_list.columns = ['user', 'rec_1','rec_2','rec_3', 'rec_4','rec_5','rec_6', 'rec_7','rec_8','rec_9', 'rec_10']
        true_list = pd.DataFrame.from_dict(y_true_list, orient="index").reset_index()
        true_list.columns = ['user', 'true_1', 'true_2', 'true_3' ,'true_4']
        
        
        print(true_list) 
        print(rec_list) 

I have noticed the following from the printed dataframe:

1. In y_true_list, there can be more than more item and in one example the item number is repeated. An example output is shown below: 9089:array(‘I’, [7, 7, 7]) 9023: array(‘I’, [17, 14]) Is it possible to have more than one item per user and have duplicate items for a specific user?

A full version of the updated function can be found below:

def print_metrics_ranking(self,df,metrics, y_prob=None, y_true=None, y_reco_list=None,
                          y_true_list=None, users=None, k=10, train=True):
    #global eval_df
    #global v
    eval_df=pd.DataFrame()
    
    if train:
        for m in metrics:
            if m in ["log_loss", "loss"]:
                log_loss_ = log_loss(y_true, y_prob, eps=1e-7)
                print(f"\t train log_loss: {log_loss_:.4f}")
                eval_df['log_loss']=[log_loss_]      
    else:
        for m in metrics:

            if m in ["log_loss", "loss"]:
                log_loss_ = log_loss(y_true, y_prob, eps=1e-7)
                print(f"\t eval log_loss: {log_loss_:.4f}")
                eval_df['log_loss_']=[log_loss_]
                
            elif m == "balanced_accuracy":
                y_pred = np.round(y_prob)
                accuracy = balanced_accuracy_score(y_true, y_pred)
                print(f"\t eval balanced accuracy: {accuracy:.4f}")
                eval_df['accuracy']=[round(accuracy,4)]
                
                
            elif m == "roc_auc":
                roc_auc = roc_auc_score(y_true, y_prob)
                print(f"\t eval roc_auc: {roc_auc:.4f}")
                eval_df['roc_auc']=[round(roc_auc,4)]

            elif m == "pr_auc":
                precision, recall, _ = precision_recall_curve(y_true, y_prob)
                pr_auc = auc(recall, precision)
                eval_df['pr_auc']=[round(pr_auc,4)]
                
            elif m == "precision":
                precision_all = precision_at_k(y_true_list, y_reco_list,
                                               users, k)
                print(f"\t eval precision@{k}: {precision_all:.4f}")
                eval_df['precision_all']=[round(precision_all,4)]
                
            elif m == "recall":
                recall_all = recall_at_k(y_true_list, y_reco_list, users, k)
                print(f"\t eval recall@{k}: {recall_all:.4f}")
                eval_df['recall_all']=[round(recall_all,4)]
                
            elif m == "map":
                map_all = map_at_k(y_true_list, y_reco_list, users, k)
                print(f"\t eval map@{k}: {map_all:.4f}")
                eval_df['map_all']=[round(map_all,4)]
                
            elif m == "ndcg":
                ndcg_all = ndcg_at_k(y_true_list, y_reco_list, users, k)
                print(f"\t eval ndcg@{k}: {ndcg_all:.4f}")
                eval_df['ndcg_all']=[round(ndcg_all,4)]
                
        #print(y_true_list)
        
        df = pd.concat([df,eval_df],ignore_index=True)
        
        #print(test_results_df)
        rec_list = pd.DataFrame.from_dict(y_reco_list, orient="index").reset_index()
        rec_list.columns = ['user', 'rec_1','rec_2','rec_3', 'rec_4','rec_5','rec_6', 'rec_7','rec_8','rec_9', 'rec_10']
        true_list = pd.DataFrame.from_dict(y_true_list, orient="index").reset_index()
        true_list.columns = ['user', 'true_1', 'true_2', 'true_3' ,'true_4', 'true_5', 'true_6', 'true_7' ,'true_8', 'true_9', 'true_10', 'true_11']
        self.rec_results  = pd.concat([self.rec_results ,rec_list],ignore_index=True)
        self.true_results  = pd.concat([self.true_results ,true_list],ignore_index=True)
        
    return df,self.true_results,self.rec_results

[massquantity]

I’ve released version 0.6.0. See the updated User Guide.

[massquantity]

y_true_list is just all of the items that an user consumed in the training data, so it is natural to contain duplicate items.