diff --git a/notebooks/PredictiveMaintenance_PerformanceMetrics.ipynb b/notebooks/PredictiveMaintenance_PerformanceMetrics.ipynb
new file mode 100644
index 0000000..a5bf242
--- /dev/null
+++ b/notebooks/PredictiveMaintenance_PerformanceMetrics.ipynb
@@ -0,0 +1,441 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "4e13863d-8db2-4384-b70c-66185cd1f51b",
+ "metadata": {},
+ "source": [
+ "### Predictive Maintenance Performance Metrics using Cohen’s Kappa, Wilcoxon T, and Confusion Matrices\n",
+ "\n",
+ "> Analyzing and evaluating predictive maintenance models using Cohen’s Kappa, Wilcoxon T test, and confusion matrices to assess performance and reliability in predictive outcomes."
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "ca5e8bc5-a7b7-457b-a695-3941e4e133d3",
+ "metadata": {},
+ "source": [
+ "##### Cohen’s Kappa for Interrater Agreement\n",
+ "\n",
+ "Cohen’s kappa measures agreement between two raters or classifiers, adjusting for agreement expected by chance. In ordered time series, it can assess how closely a model’s predicted categories align with actual labels."
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "f78fd573-9429-426c-9708-23e34a3b0dc7",
+ "metadata": {},
+ "source": [
+ "##### Wilcoxon Signed-Rank Test: Paired Comparison\n",
+ "\n",
+ "Wilcoxon tests compare paired, non-normally distributed samples. In ordered time series, it can test whether a model’s predicted ranks differ systematically from actual labels or whether two forecasting methods yield significantly different results in terms of ordinal error."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "id": "cf87587f-d904-475f-b9ed-cca5548904c4",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import warnings\n",
+ "warnings.filterwarnings('ignore')"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "id": "8f17c52b-5cfa-464c-82fd-671f798f4ab9",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "!pip install -q numpy pandas matplotlib seaborn\n",
+ "!pip install -q scipy scikit-learn"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "id": "1df98730-7a74-440c-a68e-e86e06e2d247",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import numpy as np\n",
+ "import pandas as pd\n",
+ "from sklearn.metrics import cohen_kappa_score, confusion_matrix\n",
+ "from scipy.stats import wilcoxon\n",
+ "import seaborn as sns\n",
+ "import matplotlib.pyplot as plt"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "id": "a5199587-f02b-44ac-8034-b8c566f31020",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# --- Simulate health states (0: Healthy, 1: Warning, 2: Distress) ---\n",
+ "np.random.seed(42)\n",
+ "n = 200\n",
+ "states = [1]\n",
+ "for _ in range(n - 1):\n",
+ " move = np.random.choice([-1, 0, 1], p=[0.1, 0.8, 0.1])\n",
+ " states.append(np.clip(states[-1] + move, 0, 2))\n",
+ "\n",
+ "# --- Model behavior ---\n",
+ "def simulate_model(true_states, direction):\n",
+ " shift = lambda x: np.clip(x + np.random.choice([0, direction], p=[0.8, 0.2]), 0, 2)\n",
+ " return list(map(shift, true_states))\n",
+ "\n",
+ "df = pd.DataFrame({\n",
+ " 'True': states,\n",
+ " 'Model_A': simulate_model(states, direction=-1), # Conservative\n",
+ " 'Model_B': simulate_model(states, direction=1) # Aggressive\n",
+ "})"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "id": "6a066503-d0ed-455e-b565-d5f96d067267",
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "\n",
+ "
\n",
+ " \n",
+ " \n",
+ " | \n",
+ " True | \n",
+ " Model_A | \n",
+ " Model_B | \n",
+ "
\n",
+ " \n",
+ " \n",
+ " \n",
+ " | 0 | \n",
+ " 1 | \n",
+ " 1 | \n",
+ " 1 | \n",
+ "
\n",
+ " \n",
+ " | 1 | \n",
+ " 1 | \n",
+ " 1 | \n",
+ " 1 | \n",
+ "
\n",
+ " \n",
+ " | 2 | \n",
+ " 2 | \n",
+ " 2 | \n",
+ " 2 | \n",
+ "
\n",
+ " \n",
+ " | 3 | \n",
+ " 2 | \n",
+ " 2 | \n",
+ " 2 | \n",
+ "
\n",
+ " \n",
+ " | 4 | \n",
+ " 2 | \n",
+ " 1 | \n",
+ " 2 | \n",
+ "
\n",
+ " \n",
+ "
\n",
+ "