Method process

code/dataset_benchmark.py:147–165

(self)

        os.unlink(path)

    def process(self):
        file_path = os.path.join(self.raw_dir, self.raw_file_names[0])

        data = io.loadmat(file_path)
        node_features = torch.FloatTensor(data["features"].todense())
        # remove additional dimension of length 1 in raw .mat file
        node_labels = torch.LongTensor(data["label"].squeeze())
        edge_index = []
        for relation in self.relations[self.name]:
            edge_index.append(from_scipy_sparse_matrix(
                data[relation].tocoo())[0])
        edge_index = coalesce(torch.concat(edge_index, dim=1))

        data = Data(x=node_features, edge_index=edge_index, y=node_labels)

        data = self._random_split(
            data, self.seed, self.train_size, self.val_size)
        data = data if self.pre_transform is None else self.pre_transform(data)
        self.save([data], self.processed_paths[0])

    def _random_split(self, data, seed=717, train_size=0.7, val_size=0.1):
        """split the dataset into training set, validation set and testing set"""