Created new per class accuracy function.

docs/mkdocs.yml

@@ -58,6 +58,7 @@ nav:
     - user_guide/data/three_blobs_data.md
     - user_guide/data/wine_data.md
   - evaluate:
+    - user_guide/evaluate/accuracy_score.md
     - user_guide/evaluate/bias_variance_decomp.md
     - user_guide/evaluate/bootstrap.md
     - user_guide/evaluate/bootstrap_point632_score.md

docs/sources/CHANGELOG.md

@@ -18,6 +18,7 @@ The CHANGELOG for the current development version is available at
 ##### New Features
 
 - The `SequentialFeatureSelector` now supports using pre-specified feature sets via the `fixed_features` parameter. ([#578](https://github.com/rasbt/mlxtend/pull/578))
+- Adds a new `accuracy_score` function to `mlxtend.evaluate` for computing basic classifcation accuracy, per-class accuracy, and average per-class accuracy. ([#624](https://github.com/rasbt/mlxtend/pull/624) via [Deepan Das](https://github.com/deepandas11))
 
 ##### Changes
 

docs/sources/USER_GUIDE_INDEX.md

@@ -24,6 +24,7 @@
 - [wine_data](user_guide/data/wine_data.md)
 
 ## `evaluate`
+- [accuracy_score](user_guide/evaluate/accuracy_score.md)
 - [bias_variance_decomp](user_guide/evaluate/bias_variance_decomp.md)
 - [bootstrap](user_guide/evaluate/bootstrap.md)
 - [bootstrap_point632_score](user_guide/evaluate/bootstrap_point632_score.md)

docs/sources/user_guide/evaluate/accuracy_score.ipynb

@@ -0,0 +1,327 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%matplotlib inline"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Accuracy Score"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "A function for computing for computing basic classifcation accuracy, per-class accuracy, and average per-class accuracy."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "> from mlxtend.evaluate import accuracy_score"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Example 1 -- Standard Accuracy"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The \"overall\" accuracy is defined as the number of correct predictions (*true positives* TP and *true negatives* TN) over all samples *n*:\n",
+    "\n",
+    "$$ACC = \\frac{TP + TN}{n}$$\n",
+    "\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "0.5555555555555556"
+      ]
+     },
+     "execution_count": 2,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "from mlxtend.evaluate import accuracy_score\n",
+    "\n",
+    "\n",
+    "y_targ = [0, 0, 0, 1, 1, 1, 2, 2, 2]\n",
+    "y_pred = [1, 0, 0, 0, 1, 2, 0, 2, 2]\n",
+    "\n",
+    "accuracy_score(y_targ, y_pred)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Example 2 -- Per-Class Accuracy"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The per-class accuracy is the accuracy of one class (defined as the `pos_label`) versus all remaining datapoints in the dataset."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Standard accuracy: 55.56%\n",
+      "Class 1 accuracy: 66.67%\n"
+     ]
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "from mlxtend.evaluate import accuracy_score\n",
+    "\n",
+    "\n",
+    "y_targ = [0, 0, 0, 1, 1, 1, 2, 2, 2]\n",
+    "y_pred = [1, 0, 0, 0, 1, 2, 0, 2, 2]\n",
+    "\n",
+    "std_acc = accuracy_score(y_targ, y_pred)\n",
+    "bin_acc = accuracy_score(y_targ, y_pred, method='binary', pos_label=1)\n",
+    "\n",
+    "print(f'Standard accuracy: {std_acc*100:.2f}%')\n",
+    "print(f'Class 1 accuracy: {bin_acc*100:.2f}%')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Example 3 -- Average Per-Class Accuracy"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Overview"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The \"overall\" accuracy is defined as the number of correct predictions (*true positives* TP and *true negatives* TN) over all samples *n*:\n",
+    "\n",
+    "$$ACC = \\frac{TP + TN}{n}$$\n",
+    "\n",
+    "in a binary class setting:\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "![](./scoring_files/conf_mat_binary-1.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "In a multi-class setting, we can generalize the computation of the accuracy as the fraction of all true predictions (the diagonal) over all samples n.\n",
+    "\n",
+    "$$ACC = \\frac{T}{n}$$\n",
+    "\n",
+    "Considering a multi-class problem with 3 classes (C0, C1, C2)\n",
+    "\n",
+    "![](./scoring_files/conf_mat_multi-1.png)\n",
+    "\n",
+    "let's assume our model made the following predictions:\n",
+    "\n",
+    "![](./scoring_files/multi-ex-1.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We compute the accuracy as:\n",
+    "    \n",
+    "$$ACC = \\frac{3 + 50 + 18}{90} \\approx 0.79$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now, in order to compute the ***average per-class accuracy***, we compute the binary accuracy for each class label separately; i.e., if class 1 is the positive class, class 0 and 2 are both considered the negative class.\n",
+    "\n",
+    "$$APC\\;ACC = \\frac{83/90 + 71/90 + 78/90}{3} \\approx 0.86$$"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Standard accuracy: 55.56%\n",
+      "Class 1 accuracy: 66.67%\n",
+      "Average per-class accuracy: 70.37%\n"
+     ]
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "from mlxtend.evaluate import accuracy_score\n",
+    "\n",
+    "\n",
+    "y_targ = [0, 0, 0, 1, 1, 1, 2, 2, 2]\n",
+    "y_pred = [1, 0, 0, 0, 1, 2, 0, 2, 2]\n",
+    "\n",
+    "std_acc = accuracy_score(y_targ, y_pred)\n",
+    "bin_acc = accuracy_score(y_targ, y_pred, method='binary', pos_label=1)\n",
+    "avg_acc = accuracy_score(y_targ, y_pred, method='average')\n",
+    "\n",
+    "print(f'Standard accuracy: {std_acc*100:.2f}%')\n",
+    "print(f'Class 1 accuracy: {bin_acc*100:.2f}%')\n",
+    "print(f'Average per-class accuracy: {avg_acc*100:.2f}%')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### References\n",
+    "\n",
+    "- [1] S. Raschka. [An overview of general performance metrics of binary classifier systems](http://arxiv.org/abs/1410.5330). Computing Research Repository (CoRR), abs/1410.5330, 2014.\n",
+    "- [2] Cyril Goutte and Eric Gaussier. [A probabilistic interpretation of precision, recall and f-score, with implication for evaluation](http://link.springer.com/chapter/10.1007/978-3-540-31865-1_25). In Advances in Information Retrieval, pages 345–359. Springer, 2005.\n",
+    "- [3] Brian W Matthews. [Comparison of the predicted and observed secondary structure of T4 phage lysozyme](http://www.sciencedirect.com/science/article/pii/0005279575901099). Biochimica et Biophysica Acta (BBA)- Protein Structure, 405(2):442–451, 1975."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## API"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "## accuracy_score\n",
+      "\n",
+      "*accuracy_score(y_target, y_predicted, method='standard', pos_label=1, normalize=True)*\n",
+      "\n",
+      "General accuracy function for supervised learning.\n",
+      "**Parameters**\n",
+      "\n",
+      "- `y_target` : array-like, shape=[n_values]\n",
+      "\n",
+      "    True class labels or target values.\n",
+      "\n",
+      "- `y_predicted` : array-like, shape=[n_values]\n",
+      "\n",
+      "    Predicted class labels or target values.\n",
+      "\n",
+      "- `method` : str, 'standard' by default.\n",
+      "\n",
+      "    The chosen method for accuracy computation.\n",
+      "    If set to 'standard', computes overall accuracy.\n",
+      "    If set to 'binary', computes accuracy for class pos_label.\n",
+      "    If set to 'average', computes average per class accuracy.\n",
+      "\n",
+      "- `pos_label` : str or int, 1 by default.\n",
+      "\n",
+      "    The class whose accuracy score is to be reported.\n",
+      "    Used only when `method` is set to 'binary'\n",
+      "\n",
+      "- `normalize` : bool, True by default.\n",
+      "\n",
+      "    If True, returns fraction of correctly classified samples.\n",
+      "    If False, returns number of correctly classified samples.\n",
+      "\n",
+      "**Returns**\n",
+      "\n",
+      "score: float\n",
+      "\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "with open('../../api_modules/mlxtend.evaluate/accuracy_score.md', 'r') as f:\n",
+    "    print(f.read())"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "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": "3.7.1"
+  },
+  "toc": {
+   "nav_menu": {},
+   "number_sections": true,
+   "sideBar": true,
+   "skip_h1_title": false,
+   "title_cell": "Table of Contents",
+   "title_sidebar": "Contents",
+   "toc_cell": false,
+   "toc_position": {},
+   "toc_section_display": true,
+   "toc_window_display": false
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

mlxtend/evaluate/__init__.py

@@ -26,6 +26,7 @@
 from .f_test import combined_ftest_5x2cv
 from .proportion_difference import proportion_difference
 from .bias_variance_decomp import bias_variance_decomp
+from .accuracy import accuracy_score
 
 __all__ = ["scoring", "confusion_matrix",
            "mcnemar_table", "mcnemar_tables",
@@ -37,4 +38,5 @@
            "feature_importance_permutation",
            "RandomHoldoutSplit", "PredefinedHoldoutSplit",
            "ftest", "combined_ftest_5x2cv",
-           "proportion_difference", "bias_variance_decomp"]
+           "proportion_difference", "bias_variance_decomp",
+           "accuracy_score"]