neural_network_training_template.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.utils.data import DataLoader\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import mean_squared_error, r2_score\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def remove_rows_with_missing_ratings(df):\n",
    "    df.dropna(subset=['Cleanliness_rating', 'Accuracy_rating', 'Communication_rating', 'Location_rating', 'Check-in_rating', 'Value_rating'], inplace=True)\n",
    "    return df\n",
    "  \n",
    "def combine_description_strings(df):\n",
    "    df.loc[:, 'Description'] = df['Description'].apply(lambda x: '' if pd.isna(x) else ' '.join([i for i in x if i != \"\"]).replace('About this space', '').strip())\n",
    "    df = df.dropna(subset=['Description'])\n",
    "    return df\n",
    "\n",
    "def set_default_feature_values(df):\n",
    "    df.loc[:, 'guests'] = df['guests'].apply(lambda x: 1 if pd.isna(x) else x)\n",
    "    df.loc[:, 'beds'] = df['beds'].apply(lambda x: 1 if pd.isna(x) else x)\n",
    "    df.loc[:, 'bathrooms'] = df['bathrooms'].apply(lambda x: 1 if pd.isna(x) else x)\n",
    "    df.loc[:, 'bedrooms'] = df['bedrooms'].apply(lambda x: 1 if pd.isna(x) else x)\n",
    "    return df\n",
    "\n",
    "def clean_tabular_data(df):\n",
    "    df = remove_rows_with_missing_ratings(df)\n",
    "    df = combine_description_strings(df)\n",
    "    df = set_default_feature_values(df)\n",
    "    return df\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def check_for_nan(df):\n",
    "  return df.isnull().values.any()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_airbnb(label):\n",
    "    # Load the cleaned tabular data\n",
    "    df = pd.read_csv('C:/Users/dongc/Desktop/Code/python/AiCore/airbnb_model/tabular_data/tabular_data/clean_tabular_data_new.csv')\n",
    "    \n",
    "    # Select only the numerical columns as features\n",
    "    features = df.select_dtypes(include=['float', 'int'])\n",
    "    \n",
    "    # Remove the label column from the features\n",
    "    features = features.drop(columns=[label])\n",
    "    features = features.drop(columns=['Unnamed: 19'])\n",
    "    # Select the label column as the labels\n",
    "    labels = df[label].dropna()\n",
    "    \n",
    "    return (features, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "class AirbnbNightlyPriceDataset(torch.utils.data.Dataset):\n",
    "    def __init__(self, features, labels):\n",
    "        # Convert the features and labels dataframes into torch tensors\n",
    "        self.X = torch.tensor(features.values, dtype=torch.float32)\n",
    "        self.y = torch.tensor(labels.values, dtype=torch.float32)\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        return (self.X[index], self.y[index].unsqueeze(dim=0))\n",
    "    \n",
    "    def __len__(self):\n",
    "        return len(self.X)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "class FullyConnectedNet(nn.Module):\n",
    "    def __init__(self, input_size, hidden_sizes, output_size):\n",
    "        super().__init__()\n",
    "        self.fc1 = nn.Linear(input_size, hidden_sizes[0])\n",
    "        self.fc2 = nn.Linear(hidden_sizes[0], hidden_sizes[1])\n",
    "        self.af1 = nn.ReLU()\n",
    "        self.fc3 = nn.Linear(hidden_sizes[1], output_size)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        x = self.fc1(x)\n",
    "        x = self.fc2(x)\n",
    "        x = self.af1(x)\n",
    "        x = self.fc3(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(model, dataloader, num_epochs, optimiser, criterion):\n",
    "    optimiser = optimiser\n",
    "    criterion = criterion\n",
    "    for epoch in range(10):\n",
    "        for i, (X, y) in enumerate(dataloader):\n",
    "            # Forward pass\n",
    "            try:\n",
    "                output = model(X)\n",
    "            except Exception as e:\n",
    "                print(f\"Error occurred on row {i}: {e}\")\n",
    "                continue\n",
    "            \n",
    "            loss = criterion(output, y)\n",
    "\n",
    "            # Backward pass\n",
    "            optimiser.zero_grad()\n",
    "            loss.backward()\n",
    "            optimiser.step()\n",
    "\n",
    "        # Break out of loop after first epoch\n",
    "        break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Root Mean Squared Error: 104.19483438693705\n",
      "R2 value: 0.34504199225193677\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if __name__ == \"__main__\":\n",
    "    # Load the tabular data\n",
    "    df = pd.read_csv('C:/Users/dongc/Desktop/Code/python/AiCore/airbnb_model/tabular_data/tabular_data/listing.csv')\n",
    "\n",
    "    # Clean the data\n",
    "    df = clean_tabular_data(df)\n",
    "\n",
    "    # Save the processed data\n",
    "    df.to_csv('C:/Users/dongc/Desktop/Code/python/AiCore/airbnb_model/tabular_data/tabular_data/clean_tabular_data_new.csv', index=False)\n",
    "\n",
    "    # Load the cleaned tabular data\n",
    "    features, labels = load_airbnb(label='Price_Night')\n",
    "\n",
    "    # Split the data into training and test sets\n",
    "    X_train, X_test, y_train, y_test = train_test_split(features, labels, test_size=0.2)\n",
    "    \n",
    "    # Define the dataloaders\n",
    "    batch_size = 128\n",
    "    train_dataloader = DataLoader(AirbnbNightlyPriceDataset(X_train, y_train), batch_size=batch_size, shuffle=True)\n",
    "    test_dataloader = DataLoader(AirbnbNightlyPriceDataset(X_test, y_test), batch_size=batch_size, shuffle=False)\n",
    "\n",
    "    # Define the model\n",
    "    input_size = 9\n",
    "    hidden_sizes = [5,3]\n",
    "    output_size = 1\n",
    "    model = FullyConnectedNet(input_size,hidden_sizes,output_size)\n",
    "\n",
    "    # Define the optimizer and criterion\n",
    "    optimiser = torch.optim.SGD(model.parameters(), lr=0.00002, momentum=0.7)\n",
    "    criterion = torch.nn.MSELoss()\n",
    "\n",
    "    # Train the model\n",
    "    model.train()\n",
    "    loss_accum=0\n",
    "    \n",
    "    for epoch in range(10000):\n",
    "        \n",
    "        for i, (X, y) in enumerate(train_dataloader):\n",
    "            \n",
    "            # Forward pass\n",
    "            output = model(X)\n",
    "            \n",
    "            loss = criterion(output, y)\n",
    "            loss_accum +=loss.item()\n",
    "            \n",
    "            # Backward pass\n",
    "            optimiser.zero_grad()\n",
    "            loss.backward()\n",
    "            optimiser.step()\n",
    "\n",
    "    # Test the model\n",
    "    model.eval()\n",
    "    with torch.no_grad():\n",
    "        y_pred = []\n",
    "        for X, y in test_dataloader:\n",
    "            output = model(X)\n",
    "            y_pred.append(output.numpy())\n",
    "        y_pred = np.concatenate(y_pred)\n",
    " \n",
    "    # Calculate the root mean squared error\n",
    "    rmse = np.sqrt(mean_squared_error(y_test, y_pred))\n",
    "    r2_val = r2_score(y_test, y_pred)\n",
    "    print(f\"Root Mean Squared Error: {rmse}\")\n",
    "    print(f\"R2 value: {r2_val}\")\n",
    "\n",
    "    # Log results into logs\n",
    "    output_for_log = output.detach().numpy()\n",
    "    y_for_log = y.detach().numpy()\n",
    "\n",
    "    # Convert the lists of predicted and true labels to numpy arrays\n",
    "    if isinstance(y_test,pd.Series):\n",
    "        y_test = y_test.to_numpy().flatten()\n",
    "    else:\n",
    "        y_test = y_test.flatten()\n",
    "\n",
    "    y_pred = y_pred.flatten()\n",
    "\n",
    "    # Pull out the array within the list\n",
    "\n",
    "    # Create a figure and axis\n",
    "    fig, ax = plt.subplots()\n",
    "\n",
    "    # Plot the real labels on the x-axis and the predictions on the y-axis\n",
    "    ax.scatter(y_test, y_pred, label='Test')\n",
    "\n",
    "    if len(y_for_log) == len(output_for_log):\n",
    "        output_log = output_for_log\n",
    "        ax.scatter(y_for_log, output_log, label = 'Train')\n",
    "\n",
    "    # Fit a line to the data\n",
    "    coefficients = np.polyfit(y_test, y_pred, 1)\n",
    "\n",
    "    # Create a function object for the fitted line\n",
    "    fit_fn = np.poly1d(coefficients)\n",
    "\n",
    "    # Plot the fitted line\n",
    "    ax.plot(y_test, fit_fn(y_test), '--k', label='Regression Line')\n",
    "\n",
    "    # Set the x and y axis labels\n",
    "    ax.set_xlabel('True Price per Night ($) ')\n",
    "    ax.set_ylabel('Predicted Price per Night ($)')\n",
    "    ax.text(0.05, 0.95, f\"R2 value: {r2_val:.2f}\", transform=ax.transAxes, verticalalignment='top')\n",
    "\n",
    "    # Add a legend and show the plot\n",
    "    ax.legend()\n",
    "    plt.show()\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "datascience",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.8 | packaged by conda-forge | (main, Nov 24 2022, 14:07:00) [MSC v.1916 64 bit (AMD64)]"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "6f4fdd0f09dde6c9c78e95782e94bdf3b5d4fb9a974195a1d7f1631fdbfb686e"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}