Add functionality to plot true close vs prediction on TGCN

TGCN_model.ipynb

@@ -76,7 +76,8 @@
     "from torch_geometric_temporal.nn.recurrent import A3TGCN2, DCRNN, TGCN2, TGCN,  GConvGRU, GConvLSTM, GCLSTM\n",
     "from torch.nn import Linear\n",
     "from torch.nn import ReLU\n",
-    "import matplotlib.pyplot as plt\n"
+    "import matplotlib.pyplot as plt\n",
+    "import matplotlib.dates as mdates"
    ]
   },
   {
@@ -232,7 +233,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "edge_threshold = 0\n",
+    "edge_threshold = 0.2\n",
     "\n",
     "# train_edge_weights[train_edge_weights == 1] = 0 #We exclude self edges\n",
     "train_edge_weights[train_edge_weights < edge_threshold] = 0\n",
@@ -511,10 +512,10 @@
     "            snapsot[3] = snapsot[3].squeeze(0)\n",
     "        \n",
     "        y_hat = model(snapsot[0], snapsot[1][-1], snapsot[2][-1])\n",
-    "        y_true = y_hat\n",
-    "        y_pred = snapsot[3]\n",
+    "        y_true = snapsot[3]\n",
+    "        y_pred = y_hat\n",
     "        \n",
-    "        y_pred_ar = y_hat.detach().numpy()\n",
+    "        y_pred_ar = y_pred.detach().numpy()\n",
     "        for y in y_pred_ar:\n",
     "            predictions.append(y.reshape(num_nodes))\n",
     "        \n",
@@ -764,6 +765,49 @@
     "    plt.show()"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "cce38d7f",
+   "metadata": {},
+   "source": [
+    "<font size=\"3\"> \n",
+    "This function plots the true and predicted values for a given asset over time.\n",
+    "</font>"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "44b7fdee",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def plot_precitions_vs_actuals(pivot_true,pivot_pred,asset,metrics_plot_path):\n",
+    "    # Convert datetime objects to matplotlib dates\n",
+    "    dates_true = mdates.date2num(pivot_true.index)\n",
+    "    dates_pred = mdates.date2num(pivot_pred.index)\n",
+    "\n",
+    "    # Plotting the true values\n",
+    "    plt.plot(dates_true, pivot_true[asset], label='True Values')\n",
+    "    # Plotting the predicted values\n",
+    "    plt.plot(dates_pred, pivot_pred[asset], label='Predictions')\n",
+    "\n",
+    "    # Formatting the x-axis dates\n",
+    "    plt.gca().xaxis.set_major_formatter(mdates.DateFormatter('%m/%d/%Y'))\n",
+    "    plt.gca().xaxis.set_major_locator(mdates.DayLocator(interval=20)) # set tick interval to 30 days\n",
+    "\n",
+    "    # Adding labels and title\n",
+    "    plt.xlabel('Date')\n",
+    "    plt.ylabel('Close')\n",
+    "    plt.title(asset + ': Closes vs Predictions')\n",
+    "\n",
+    "    # Adding legend\n",
+    "    plt.legend()\n",
+    "    plt.savefig(metrics_plot_path+asset+'_True_vs_Pred.pdf')\n",
+    "    # Displaying the plot\n",
+    "    plt.show()"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "8bcbf004",
@@ -896,7 +940,7 @@
     "train_edge_weights_f = np.concatenate((train_edge_weights, val_edge_weights), axis=0)\n",
     "train_node_features_f = np.concatenate((train_node_features, val_node_features), axis=0)\n",
     "train_node_labels_f = np.concatenate((train_node_labels, val_node_labels), axis=0)\n",
-    "epochs = 30\n",
+    "epochs = 15\n",
     "\n",
     "train_loader,test_loader = graph_data_loader(batch_size, train_edges_f, train_edge_weights_f, train_node_features_f,\n",
     "                                train_node_labels_f, test_edges, test_edge_weights, test_node_features, test_node_labels)\n",
@@ -945,6 +989,24 @@
     "plot_single_metric(metrics_test['eval_mae_ls'],'MAE','Test',metrics_plot_path,gcn)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "53ca301d",
+   "metadata": {},
+   "source": [
+    "### Plot Predicted and True Closes"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "dd220e76",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "plot_precitions_vs_actuals(pivot_true,pivot_pred,'BTC',metrics_plot_path)"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "f08f199a",
@@ -981,9 +1043,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "epxeriment_name = 'Final_Proposed'\n",
-    "with open(results_path+gcn+'_'+epxeriment_name+'.txt', \"w\") as file:\n",
-    "    file.write(str(metrics_test['eval_mae_ls']))"
+    "# epxeriment_name = 'Final_Proposed'\n",
+    "# with open(results_path+gcn+'_'+epxeriment_name+'.txt', \"w\") as file:\n",
+    "#     file.write(str(metrics_test['eval_mae_ls']))"
    ]
   },
   {
@@ -1042,14 +1104,6 @@
     "plt.savefig(results_path+\"Edges_Infulence.pdf\")\n",
     "plt.show()"
    ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "1c186bb6",
-   "metadata": {},
-   "outputs": [],
-   "source": []
   }
  ],
  "metadata": {