AMUGraph/graph_dataset.py at main · agarwalishika/AMUGraph · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
import pandas as pd
import torch
from torch.utils.data import Dataset
import pickle
import networkx as nx
import os
from tqdm import tqdm
import sqlite3
import numpy as np
import random
import dgl
from dgl.data import FraudYelpDataset

class AMUGraphDataset(Dataset):
    def __init__(self, dataset="", train=True, dataset_label=-1):
        device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
        self.dataset = dataset

        print(f'Reading from {self.dataset}...')

        meta = pd.read_csv(f'{self.dataset}_embeddings/final_embeddings0.csv', sep=',')

        for i in range(1,20):
            m = pd.read_csv(f'{self.dataset}_embeddings/final_embeddings{i}.csv', sep=',')
            meta = pd.concat([meta, m])

        meta = meta.sample(frac=1).reset_index(drop=True)

        '''meta = pd.read_csv("ecg_final.txt", sep='  ', header=None)
        meta = meta.add_prefix('c')
        label = meta.pop('c0')
        label = [1 if x >= 3 else 0 for x in label]
        meta.insert(140, 'c0', label)'''

        train_size = int(0.75 * len(meta))
        test_size = int(0.25 * len(meta))
        self.features = meta.iloc[:, 1:-1]


        if train:
            features = meta.iloc[:train_size, 1:-1]
            labels = meta.iloc[:train_size, -1]
        else:
            features = meta.iloc[train_size:train_size+test_size, 1:-1]
            labels = meta.iloc[train_size:train_size+test_size, -1]

        self.node_to_str = meta.iloc[:, 0].reset_index(drop=True)

        self.str_to_node = pd.DataFrame(self.node_to_str.copy())
        self.str_to_node.columns = ['name']
        self.str_to_node = self.str_to_node.reset_index()
        self.str_to_node = self.str_to_node.set_index('name')

        self.k = 2

        self.x = torch.tensor(features.values, dtype=torch.float32).to(device)
        self.y = torch.tensor(labels.values).to(device)

        if dataset_label >= 0:
            ind = torch.nonzero(self.y==dataset_label).reshape(-1)
            self.y = self.y[ind]
            self.x = self.x[ind]


        # if not os.path.exists(f'{self.dataset}neighbors.txt'):
        #     self.print_neighbors()

    def __len__(self):
        return len(self.y)

    def __getitem__(self, index):
        index = int(index)
        return self.x[index], self.y[index], index

    def get_neighbors(self, node):
        conn = sqlite3.connect(f'{self.dataset}neighbors.db')
        cursor = conn.cursor()
        cursor.execute(f'SELECT neighbors FROM neighbors WHERE node_num = {node}')
        result = cursor.fetchone()
        conn.close()

        neighbors = result[0]
        neighbors = neighbors.split(',')
        neighbors = [int(i) for i in neighbors if i != '']
        neighbors = [i for i in neighbors if i < self.features.shape[0]]

        # for debugging purposes
        with open('progress.txt', 'a') as f:
            f.write(f'{node}, ')

        return neighbors

    def get_random_fast(self, neighbors):
        do_not_pick = {}
        for n in neighbors:
            do_not_pick[n] = 1

        random_nodes = []
        while len(random_nodes) != len(neighbors):
            x = random.randint(0,self.features.shape[0] - 1)
            try:
                do_not_pick[x]
            except KeyError:
                random_nodes.append(x)

        return random_nodes

    def get_random(self, node, num_nodes):
        conn = sqlite3.connect(f'{self.dataset}random.db')
        cursor = conn.cursor()
        cursor.execute(f'SELECT neighbors FROM random_neighbors WHERE node_num = {node}')

        result = cursor.fetchone()
        conn.close()

        nodes = result[0]
        nodes = nodes.split(',')
        nodes = [int(i) for i in nodes if i != '']

        return np.random.choice(nodes, num_nodes)
        '''while len(samples) < num_nodes:
            samp = self.node_to_str.sample().index[0]
            if samp not in samples or samp not in neighbors:
                samples.append(samp)'''

        return samples

    def print_neighbors(self):
        print('Loading all the neighbors')
        '''f = open(f'{self.dataset}graph.txt', 'rb')
        G = pickle.load(f)
        f.close()'''

        dataset = FraudYelpDataset()
        g = dataset[0]
        G = dgl.to_homogeneous(g)
        G = dgl.to_networkx(G)

        nodes = G.nodes
        for n in tqdm(nodes):
            neighbors = nx.single_source_shortest_path_length(G, n, self.k)
            neighbors = [k for k,v in neighbors.items() if n is not k] #[self.str_to_node.loc[i][0] for i in neighbors if i is not n]
            neighbors = ','.join(str(n) for n in neighbors)
            f = open(f'{self.dataset}neighbors.txt', 'a')
            f.write(f'{n}: {neighbors}\n')
            f.close()
        print('Done loading all the neighbors into neighbors.txt')
        return neighbors