Risk Commodity Dataset (RCDD) from Alibaba's e-commerce platform

<p>This is a risk commodity detection dataset (RCDD) that is based on a real risk detection scenario from Alibaba&#39;s e-commerce platform. All the <em>.csv&nbsp;</em>files are the initial data which&nbsp;consists of edges, node features, supervised information as well as IDs of candidate items. More details can see in <em>README.md</em>, we list the type of each column in each file as follows:</p> <p>1. RCDD_edges.csv (edge file):&nbsp;<br> &nbsp; &nbsp; source_node_id int, target_node_id int, source_node_type string, target_node_type string, edge_type string<br> 2. RCDD_nodes.csv (node file):<br> &nbsp; &nbsp; node_id int, node_type string, node_atts string (notice that node_atts are 256-dimensional feature vector strings with delimiter &quot;:&quot;)<br> 3. RCDD_train_labels.csv (training labels):<br> &nbsp; &nbsp; item_id int, label int<br> 4. RCDD_test_ids.csv (testing ids):<br> &nbsp; &nbsp; item_id int<br> 5. RCDD_test_labels.csv (testing labels):<br> &nbsp; &nbsp; item_id int, label int&nbsp; &nbsp;&nbsp;</p> <p>Besides that,<em> graph.bin</em> is the format in DGL which is constructed by&nbsp;&nbsp;all&nbsp;<em>*.csv</em>&nbsp;files, and a&nbsp;general method to load this graph as follows:</p> <pre><code class="language-python">from dgl import load_graphs #should install dgl ds,_ = load_graphs("./graph.bin") g = ds[0] print(g)</code></pre> <p>And then you can easily get a large-scale heterogeneous graph with 157,814,864 edges and 13,806,619 nodes, our graph task is node classification: detect risk product.</p>