Update 04_preprocessing_and_training_LinaAbdullahi.ipynb

Notebooks/04_preprocessing_and_training_LinaAbdullahi.ipynb

@@ -3988,7 +3988,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 110,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -4013,9 +4013,18 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 111,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Directory ../models was created.\n",
+      "Writing file.  \"../models\\ski_resort_pricing_model.pkl\"\n"
+     ]
+    }
+   ],
    "source": [
     "# save the model\n",
     "\n",
@@ -4050,14 +4059,11 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "In this preprocessing step, we first evaluated the usefulness of the mean value as a predictor. Then, constructed a data processing pipeline that imputed missing values, scaled the data, selected the most relevant features using the `select k best`strategy, and trained a linear regression model while employing cross-validation to estimate its performance.\n",
-    "\n",
-    "In our preprocessing, we later compared imputing missing values with both the median and mean approaches and assessed their performance using a linear regression model. Interestingly, the choice of imputation method (median or mean) didn't have a significant impact on the model's accuracy. On average, the model could estimate a ticket price within approximately 9 dollars of the actual price, which was much better than a simple guess based on the average value (with a variance of 19 dollars).\n",
+    "In this preprocessing step, we first evaluated the usefulness of the mean value as a predictor. we then compared imputing missing values with both the median and mean approaches and assessed their performance using a linear regression model. Interestingly, the choice of imputation method (median or mean) didn't have a significant impact on the model's accuracy. On average, the model could estimate a ticket price within approximately 9 dollars of the actual price, which was much better than a simple guess based on the average value (with a variance of 19 dollars).\n",
     "\n",
     "Next, we employed a pipeline that involved imputing missing values, scaling the data, and performing linear regression in a single process. To identify the most influential/dominant features, we utilized the `select k best` strategy. After experimenting with setting different values of K, we did a hyperparameter search using GridSearchCV and found that 8 was the best value for k (8 number of features). The top 8 features that the linear regression model identified as the most important for predicting ticket prices were: `vertical_drop`, `Snow Making_ac`, `total_chairs`, `fastQuads`, `Runs`, `LongestRun_mi`, `trams`, and `SkiableTerrain_ac`.\n",
     "\n",
     "Subsequently, we explored the random forest model using cross-validation and found that it aligned with the linear model, highlighting the importance of four features: `fastQuads`, `Runs`, `Snow Making_ac`, and `vertical_drop`.\n",
-    "\n",
     "Comparing the performance of the `linear regression` and `random forest models`, we observed that the random forest model exhibited a lower cross-validation mean absolute error, almost 1 dollar less. As a result, we conclude that the preferred choice for this project is the `random forest model`, selected for its superior stability and lower cross-validation mean absolute error.\n",
     "\n",
     "Validating the performance on the test set yielded consistent results, affirming the reliability and effectiveness of the random forest model for predicting ticket prices."