Day_43_Regression_using_TensorFlow.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyOgDJEPspt/3jNHAgnNeVDg"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "!pip install utils"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Xe9P79LBVZKn",
        "outputId": "56697b78-aa01-4d43-8de2-8ad7c04a2968"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting utils\n",
            "  Downloading utils-1.0.1-py2.py3-none-any.whl (21 kB)\n",
            "Installing collected packages: utils\n",
            "Successfully installed utils-1.0.1\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "POMyuM8nRrt4"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import tensorflow as tf\n",
        "from utils import *\n",
        "from sklearn.model_selection import train_test_split\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Dense, Dropout\n",
        "from tensorflow.keras.callbacks import EarlyStopping, LambdaCallback\n",
        "%matplotlib inline\n",
        "tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Load the Data"
      ],
      "metadata": {
        "id": "JKcgj3oaR2QW"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "column_names = ['serial', 'date', 'distance', 'stores', 'latitude', 'longitude', 'price']\n",
        "df = pd.read_csv('https://github.com/raw/thunderstroke325/60-Days-of-Data-Science-and-ML/main/datasets/data43.csv', names =column_names)\n",
        "# check missing data\n",
        "df.isna().sum()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ijua4HXnR405",
        "outputId": "88d5f1a4-093c-42b9-fdfa-abb5bedae676"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "serial       0\n",
              "date         0\n",
              "distance     0\n",
              "stores       0\n",
              "latitude     0\n",
              "longitude    0\n",
              "price        0\n",
              "dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Normalize Data and convert label value"
      ],
      "metadata": {
        "id": "Ovbc72DdR2OK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "df = df.iloc[:,1:]\n",
        "dn = (df-df.mean())/df.std()\n",
        "y_mean = df['price'].mean()\n",
        "y_std = df['price'].std()\n",
        "def cl(pred):\n",
        "    return int(pred * y_std + y_mean)"
      ],
      "metadata": {
        "id": "6Ac-3rvSR-gR"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Select Features and Train and Test split"
      ],
      "metadata": {
        "id": "iH2iwDDHR2L8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = dn.iloc[:,:6]\n",
        "y = dn.iloc[:,-1]\n",
        "x_arr = x.values\n",
        "y_arr = y.values\n",
        "x_train,x_test,y_train,y_test = train_test_split(x_arr,y_arr,test_size =0.05,random_state=0)"
      ],
      "metadata": {
        "id": "f-OkxhorSC84"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Create the model"
      ],
      "metadata": {
        "id": "QJaPfCJxSDmQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def gm():\n",
        "    model = Sequential([\n",
        "        Dense(10,input_shape = (6,),activation='relu'),\n",
        "        Dense(20,activation = 'relu'),\n",
        "        Dense(5,activation='relu'),\n",
        "        Dense(1)\n",
        "        \n",
        "    ])\n",
        "    model.compile(\n",
        "        loss ='mse',\n",
        "        optimizer ='adam'\n",
        "        \n",
        "    )\n",
        "    return model\n",
        "gm().summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "F0Aom4aHSGaT",
        "outputId": "2ae7032c-90da-41d8-a1ec-d9c06b530d9e"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense (Dense)               (None, 10)                70        \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 20)                220       \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 5)                 105       \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 1)                 6         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 401\n",
            "Trainable params: 401\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Model Training"
      ],
      "metadata": {
        "id": "phUYTDWvSDjz"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "es = EarlyStopping(monitor = 'val_loss', patience =5)\n",
        "model =gm()\n",
        "pu = model.predict(x_test)\n",
        "h = model.fit(\n",
        "    x_train, y_train,\n",
        "    validation_data =(x_test,y_test),\n",
        "    epochs=100,\n",
        "    callbacks = [es]\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CTrCfzeBSLO6",
        "outputId": "c819abe5-1bdc-4518-e4d3-ebebb153332c"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "149/149 [==============================] - 1s 5ms/step - loss: 0.4915 - val_loss: 0.1024\n",
            "Epoch 2/100\n",
            "149/149 [==============================] - 1s 4ms/step - loss: 0.0610 - val_loss: 0.0308\n",
            "Epoch 3/100\n",
            "149/149 [==============================] - 1s 3ms/step - loss: 0.0251 - val_loss: 0.0164\n",
            "Epoch 4/100\n",
            "149/149 [==============================] - 1s 4ms/step - loss: 0.0134 - val_loss: 0.0106\n",
            "Epoch 5/100\n",
            "149/149 [==============================] - 1s 4ms/step - loss: 0.0081 - val_loss: 0.0071\n",
            "Epoch 6/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0054 - val_loss: 0.0052\n",
            "Epoch 7/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0041 - val_loss: 0.0041\n",
            "Epoch 8/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0032 - val_loss: 0.0033\n",
            "Epoch 9/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0027 - val_loss: 0.0028\n",
            "Epoch 10/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0023 - val_loss: 0.0023\n",
            "Epoch 11/100\n",
            "149/149 [==============================] - 0s 3ms/step - loss: 0.0019 - val_loss: 0.0021\n",
            "Epoch 12/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0017 - val_loss: 0.0017\n",
            "Epoch 13/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0015 - val_loss: 0.0016\n",
            "Epoch 14/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0013 - val_loss: 0.0013\n",
            "Epoch 15/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0011 - val_loss: 0.0012\n",
            "Epoch 16/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 0.0010 - val_loss: 0.0010\n",
            "Epoch 17/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 8.8489e-04 - val_loss: 9.1113e-04\n",
            "Epoch 18/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 7.9933e-04 - val_loss: 8.6668e-04\n",
            "Epoch 19/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 7.1532e-04 - val_loss: 7.4594e-04\n",
            "Epoch 20/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 6.5376e-04 - val_loss: 6.6500e-04\n",
            "Epoch 21/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 5.8485e-04 - val_loss: 7.2834e-04\n",
            "Epoch 22/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 5.3075e-04 - val_loss: 6.0480e-04\n",
            "Epoch 23/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.7009e-04 - val_loss: 4.8716e-04\n",
            "Epoch 24/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.2593e-04 - val_loss: 4.7348e-04\n",
            "Epoch 25/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 3.9264e-04 - val_loss: 4.2426e-04\n",
            "Epoch 26/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 3.5851e-04 - val_loss: 3.5444e-04\n",
            "Epoch 27/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 3.2225e-04 - val_loss: 3.2594e-04\n",
            "Epoch 28/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.9298e-04 - val_loss: 3.0120e-04\n",
            "Epoch 29/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.7086e-04 - val_loss: 3.1783e-04\n",
            "Epoch 30/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.5093e-04 - val_loss: 2.6940e-04\n",
            "Epoch 31/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.3794e-04 - val_loss: 2.2285e-04\n",
            "Epoch 32/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.0742e-04 - val_loss: 2.0535e-04\n",
            "Epoch 33/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.7876e-04 - val_loss: 1.9340e-04\n",
            "Epoch 34/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.6165e-04 - val_loss: 1.7644e-04\n",
            "Epoch 35/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.5093e-04 - val_loss: 1.7567e-04\n",
            "Epoch 36/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.3787e-04 - val_loss: 1.6351e-04\n",
            "Epoch 37/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.2415e-04 - val_loss: 1.6389e-04\n",
            "Epoch 38/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.1229e-04 - val_loss: 1.1524e-04\n",
            "Epoch 39/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.0271e-04 - val_loss: 1.2357e-04\n",
            "Epoch 40/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 8.8340e-05 - val_loss: 1.0765e-04\n",
            "Epoch 41/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 7.9756e-05 - val_loss: 1.1816e-04\n",
            "Epoch 42/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 7.5189e-05 - val_loss: 1.0239e-04\n",
            "Epoch 43/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 7.3513e-05 - val_loss: 6.7512e-05\n",
            "Epoch 44/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 5.7674e-05 - val_loss: 7.5165e-05\n",
            "Epoch 45/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 5.5083e-05 - val_loss: 8.5737e-05\n",
            "Epoch 46/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.9845e-05 - val_loss: 6.2448e-05\n",
            "Epoch 47/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.4433e-05 - val_loss: 6.9165e-05\n",
            "Epoch 48/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.9399e-05 - val_loss: 5.4599e-05\n",
            "Epoch 49/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.0966e-05 - val_loss: 5.2438e-05\n",
            "Epoch 50/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 3.8776e-05 - val_loss: 5.7642e-05\n",
            "Epoch 51/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 4.1892e-05 - val_loss: 4.2101e-05\n",
            "Epoch 52/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 3.1405e-05 - val_loss: 2.6106e-05\n",
            "Epoch 53/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.8029e-05 - val_loss: 4.4550e-05\n",
            "Epoch 54/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.9333e-05 - val_loss: 3.8637e-05\n",
            "Epoch 55/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.3751e-05 - val_loss: 2.4844e-05\n",
            "Epoch 56/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.1793e-05 - val_loss: 2.0407e-05\n",
            "Epoch 57/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.4664e-05 - val_loss: 2.3528e-05\n",
            "Epoch 58/100\n",
            "149/149 [==============================] - 0s 3ms/step - loss: 1.9955e-05 - val_loss: 2.4583e-05\n",
            "Epoch 59/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.7602e-05 - val_loss: 2.2754e-05\n",
            "Epoch 60/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.4347e-05 - val_loss: 2.6970e-05\n",
            "Epoch 61/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.0621e-05 - val_loss: 1.6926e-05\n",
            "Epoch 62/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.5620e-05 - val_loss: 1.5586e-05\n",
            "Epoch 63/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 2.4797e-05 - val_loss: 4.2284e-05\n",
            "Epoch 64/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.9155e-05 - val_loss: 2.4927e-05\n",
            "Epoch 65/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.8940e-05 - val_loss: 2.5621e-05\n",
            "Epoch 66/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.2857e-05 - val_loss: 1.7647e-05\n",
            "Epoch 67/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.0584e-05 - val_loss: 1.0675e-05\n",
            "Epoch 68/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.4315e-05 - val_loss: 1.6100e-05\n",
            "Epoch 69/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.0253e-05 - val_loss: 2.0210e-05\n",
            "Epoch 70/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.4496e-05 - val_loss: 6.9053e-05\n",
            "Epoch 71/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.5858e-05 - val_loss: 1.6773e-05\n",
            "Epoch 72/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 9.7922e-06 - val_loss: 7.9213e-06\n",
            "Epoch 73/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.1045e-05 - val_loss: 8.3531e-06\n",
            "Epoch 74/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 7.1072e-06 - val_loss: 6.5941e-06\n",
            "Epoch 75/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 9.9183e-06 - val_loss: 3.6126e-05\n",
            "Epoch 76/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.9992e-05 - val_loss: 8.3169e-06\n",
            "Epoch 77/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.1436e-05 - val_loss: 3.4920e-05\n",
            "Epoch 78/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 1.0207e-05 - val_loss: 1.7413e-05\n",
            "Epoch 79/100\n",
            "149/149 [==============================] - 0s 2ms/step - loss: 5.5641e-06 - val_loss: 8.0371e-06\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Plot Loss"
      ],
      "metadata": {
        "id": "G_Kgx1tgSDhR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "loss_values = h.history['loss']\n",
        "epochs = range(1, len(loss_values)+1)\n",
        "\n",
        "plt.plot(epochs, loss_values, label='Training Loss')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "tanUaDQzSN1a",
        "outputId": "9e698d86-6d35-4f4c-d791-e5d29bfeb6d2"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}