c1m1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "from sklearn import preprocessing\n",
    "\n",
    "import keras\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Input, Dense, LSTM, Dropout, Activation\n",
    "from keras.layers import Conv1D, MaxPooling1D, Flatten, BatchNormalization, Lambda, AveragePooling1D\n",
    "from keras.layers import GlobalMaxPooling1D, GlobalAveragePooling1D, Add\n",
    "from keras.callbacks import EarlyStopping\n",
    "from keras.optimizers import Adam, RMSprop, SGD\n",
    "from keras import regularizers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load NumPy Arrays & convert labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "segs = 512"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load compressed NumPy arrays\n",
    "with np.load('orpheusC' + str(segs) + '.npz') as data:\n",
    "    X_Train = data['Xt_Train']\n",
    "    X_Val = data['Xt_Val']\n",
    "    X_Test = data['Xt_Test']\n",
    "    Y_Train = data['Y_Train']\n",
    "    Y_Val  = data['Y_Val']\n",
    "    Y_Test = data['Y_Test']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Blues', 'Classical', 'Country', 'Hip-Hop', 'Jazz', 'Metal', 'Pop',\n",
       "       'Rock'], dtype='<U9')"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Convert numerical labels to integers\n",
    "le = preprocessing.LabelEncoder()\n",
    "Y_Train = le.fit_transform(Y_Train)\n",
    "Y_Val = le.fit_transform(Y_Val)\n",
    "Y_Test = le.fit_transform(Y_Test)\n",
    "le.classes_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Metal        880\n",
       " Jazz         880\n",
       " Classical    880\n",
       " Country      880\n",
       " Rock         880\n",
       " Blues        880\n",
       " Hip-Hop      880\n",
       " Pop          880\n",
       " Name: 0, dtype: int64,\n",
       " Rock         110\n",
       " Classical    110\n",
       " Country      110\n",
       " Jazz         110\n",
       " Hip-Hop      110\n",
       " Pop          110\n",
       " Metal        110\n",
       " Blues        110\n",
       " Name: 0, dtype: int64,\n",
       " Rock         110\n",
       " Classical    110\n",
       " Country      110\n",
       " Jazz         110\n",
       " Hip-Hop      110\n",
       " Pop          110\n",
       " Metal        110\n",
       " Blues        110\n",
       " Name: 0, dtype: int64)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Verify sample balance before training\n",
    "import pandas as pd\n",
    "Y_Traindf = pd.DataFrame(le.inverse_transform(Y_Train))\n",
    "Y_Valdf = pd.DataFrame(le.inverse_transform(Y_Val))\n",
    "Y_Testdf = pd.DataFrame(le.inverse_transform(Y_Test))\n",
    "Y_Traindf[0].value_counts(),Y_Valdf[0].value_counts(),Y_Testdf[0].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Build Model(s)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Specifying input shape for first layer\n",
    "input_shape = (segs, 12)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### CNN (Timbre)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set dropout rate\n",
    "dr = 0.25"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /Users/abuchmueller/opt/anaconda3/envs/P37/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:4070: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Build model using Sequential API\n",
    "model = Sequential()\n",
    "\n",
    "# 3x 1D Convolution layers\n",
    "model.add(Conv1D(48, 5, activation='relu', input_shape=input_shape, data_format='channels_last'))\n",
    "model.add(MaxPooling1D(pool_size=2))\n",
    "model.add(Dropout(rate=dr))\n",
    "\n",
    "model.add(Conv1D(64, 5, activation='relu', padding=\"same\"))\n",
    "model.add(MaxPooling1D(pool_size=2))\n",
    "model.add(Dropout(rate=dr))\n",
    "\n",
    "model.add(Conv1D(64, 5, activation='relu', padding=\"same\"))\n",
    "model.add(MaxPooling1D(pool_size=2))\n",
    "model.add(Dropout(rate=dr))\n",
    "\n",
    "model.add(Flatten()) \n",
    "model.add(Dropout(rate=dr))\n",
    "\n",
    "# Dense Layer with ReLU activation\n",
    "model.add(Dense(64, activation='relu'))\n",
    "\n",
    "# Output layer\n",
    "model.add(Dense(8, activation='softmax'))\n",
    "\n",
    "mtype = 'cnn'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Compile Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, min_delta=0.01, patience=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /Users/abuchmueller/opt/anaconda3/envs/P37/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:422: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.\n",
      "\n",
      "Train on 7040 samples, validate on 880 samples\n",
      "Epoch 1/100\n",
      "7040/7040 [==============================] - 19s 3ms/step - loss: 4.0477 - accuracy: 0.2561 - val_loss: 1.5373 - val_accuracy: 0.4114\n",
      "Epoch 2/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 1.4748 - accuracy: 0.4271 - val_loss: 1.3267 - val_accuracy: 0.5011\n",
      "Epoch 3/100\n",
      "7040/7040 [==============================] - 18s 2ms/step - loss: 1.3095 - accuracy: 0.4964 - val_loss: 1.2673 - val_accuracy: 0.5023\n",
      "Epoch 4/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 1.1996 - accuracy: 0.5378 - val_loss: 1.2406 - val_accuracy: 0.5398\n",
      "Epoch 5/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 1.1257 - accuracy: 0.5626 - val_loss: 1.1417 - val_accuracy: 0.5830\n",
      "Epoch 6/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 1.0728 - accuracy: 0.5919 - val_loss: 1.0740 - val_accuracy: 0.6068\n",
      "Epoch 7/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 1.0274 - accuracy: 0.6038 - val_loss: 1.0396 - val_accuracy: 0.6182\n",
      "Epoch 8/100\n",
      "7040/7040 [==============================] - 18s 2ms/step - loss: 0.9915 - accuracy: 0.6168 - val_loss: 1.0562 - val_accuracy: 0.6148\n",
      "Epoch 9/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.9573 - accuracy: 0.6278 - val_loss: 1.0200 - val_accuracy: 0.6250\n",
      "Epoch 10/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.9193 - accuracy: 0.6445 - val_loss: 0.9945 - val_accuracy: 0.6125\n",
      "Epoch 11/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.8914 - accuracy: 0.6601 - val_loss: 0.9748 - val_accuracy: 0.6432\n",
      "Epoch 12/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.8577 - accuracy: 0.6695 - val_loss: 0.9277 - val_accuracy: 0.6614\n",
      "Epoch 13/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.8349 - accuracy: 0.6838 - val_loss: 0.9337 - val_accuracy: 0.6614\n",
      "Epoch 14/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.8336 - accuracy: 0.6812 - val_loss: 0.9278 - val_accuracy: 0.6545\n",
      "Epoch 15/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7821 - accuracy: 0.7010 - val_loss: 0.8596 - val_accuracy: 0.6818\n",
      "Epoch 16/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7676 - accuracy: 0.7057 - val_loss: 0.8763 - val_accuracy: 0.6750\n",
      "Epoch 17/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7578 - accuracy: 0.7098 - val_loss: 0.8718 - val_accuracy: 0.6648\n",
      "Epoch 18/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7355 - accuracy: 0.7153 - val_loss: 0.8674 - val_accuracy: 0.6886\n",
      "Epoch 19/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7207 - accuracy: 0.7259 - val_loss: 0.8728 - val_accuracy: 0.6943\n",
      "Epoch 20/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7118 - accuracy: 0.7227 - val_loss: 0.8719 - val_accuracy: 0.6636\n",
      "Epoch 21/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.7021 - accuracy: 0.7276 - val_loss: 0.8600 - val_accuracy: 0.6852\n",
      "Epoch 22/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.6812 - accuracy: 0.7317 - val_loss: 0.8556 - val_accuracy: 0.6920\n",
      "Epoch 23/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.6674 - accuracy: 0.7457 - val_loss: 0.9125 - val_accuracy: 0.6705\n",
      "Epoch 24/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.6435 - accuracy: 0.7567 - val_loss: 0.8678 - val_accuracy: 0.6795\n",
      "Epoch 25/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.6504 - accuracy: 0.7484 - val_loss: 0.8747 - val_accuracy: 0.7136\n",
      "Epoch 26/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.6234 - accuracy: 0.7632 - val_loss: 0.8148 - val_accuracy: 0.6966\n",
      "Epoch 27/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.6040 - accuracy: 0.7639 - val_loss: 0.8527 - val_accuracy: 0.7011\n",
      "Epoch 28/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.5990 - accuracy: 0.7754 - val_loss: 0.8470 - val_accuracy: 0.7114\n",
      "Epoch 29/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.5907 - accuracy: 0.7756 - val_loss: 0.8584 - val_accuracy: 0.7000\n",
      "Epoch 30/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.5858 - accuracy: 0.7832 - val_loss: 0.8659 - val_accuracy: 0.6841\n",
      "Epoch 31/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.5807 - accuracy: 0.7835 - val_loss: 0.8977 - val_accuracy: 0.6886\n",
      "Epoch 32/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.5593 - accuracy: 0.7902 - val_loss: 0.8831 - val_accuracy: 0.7136\n",
      "Epoch 33/100\n",
      "7040/7040 [==============================] - 18s 3ms/step - loss: 0.5686 - accuracy: 0.7847 - val_loss: 1.0067 - val_accuracy: 0.6614\n",
      "Epoch 00033: early stopping\n"
     ]
    }
   ],
   "source": [
    "# iterating 30 times on the data in batches of 32 samples then saving training history\n",
    "hist = model.fit(X_Train, Y_Train, epochs=100, batch_size=32, validation_data=(X_Val, Y_Val), callbacks=[es])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv1d_1 (Conv1D)            (None, 508, 48)           2928      \n",
      "_________________________________________________________________\n",
      "max_pooling1d_1 (MaxPooling1 (None, 254, 48)           0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 254, 48)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_2 (Conv1D)            (None, 254, 64)           15424     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_2 (MaxPooling1 (None, 127, 64)           0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 127, 64)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_3 (Conv1D)            (None, 127, 64)           20544     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_3 (MaxPooling1 (None, 63, 64)            0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 63, 64)            0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 4032)              0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 4032)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 64)                258112    \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 8)                 520       \n",
      "=================================================================\n",
      "Total params: 297,528\n",
      "Trainable params: 297,528\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Model Summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "880/880 [==============================] - 1s 827us/step\n"
     ]
    }
   ],
   "source": [
    "model.save(os.getcwd() + '/models/timbre/' + mtype + str(segs) + '_' + str(round(model.evaluate(X_Test, Y_Test)[1], 3)) + '.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plotting training and validation loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(6,6))\n",
    "loss = hist.history['loss']\n",
    "val_loss = hist.history['val_loss']\n",
    "epochs = range(1, len(loss) + 1)\n",
    "plt.plot(epochs, loss, 'b', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'r', label='Validation loss')\n",
    "plt.axvline(x = (np.argmin(val_loss)+1))\n",
    "plt.title('Training and validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plotting training and validation accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(6,6))\n",
    "acc = hist.history['accuracy']\n",
    "val_acc = hist.history['val_accuracy']\n",
    "plt.plot(epochs, acc, 'b', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'r', label='Validation acc')\n",
    "plt.axvline(x = (np.argmax(val_acc)+1))\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(25, 0.7136363387107849)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Epoch with highest validation accucary\n",
    "np.argmax(val_acc)+1, val_acc[np.argmax(val_acc)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "880/880 [==============================] - 1s 822us/step\n",
      "Test loss: 0.9529982415112582\n",
      "Test accuracy: 0.6818181872367859\n"
     ]
    }
   ],
   "source": [
    "results = model.evaluate(X_Test, Y_Test)\n",
    "print('Test loss:', results[0])\n",
    "print('Test accuracy:', results[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Accuracy per Genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 76   0  12   0  10   2   0  10]\n",
      " [  3 104   0   0   3   0   0   0]\n",
      " [ 16   1  86   0   2   1   3   1]\n",
      " [  3   1   3  80  10   0  13   0]\n",
      " [  9  11   4   2  83   1   0   0]\n",
      " [  3   3   1   0   1  88   0  14]\n",
      " [  5   1  31  13   7   3  39  11]\n",
      " [ 25   0  13   1   5  20   2  44]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "y_pred = model.predict(X_Test)\n",
    "y_pred = np.argmax(y_pred, axis=1)\n",
    "print(confusion_matrix(Y_Test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(258.72, 0.5, 'Prediction')"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualizing confusion matrix\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import seaborn as sns\n",
    "\n",
    "plt.figure(figsize=(12,6))\n",
    "cm = confusion_matrix(Y_Test, y_pred)\n",
    "sns.heatmap(cm.T, square=True, annot=True, fmt='d', cbar=False,\n",
    "            xticklabels=list(le.classes_),\n",
    "            yticklabels=list(le.classes_))\n",
    "plt.xlabel('True Genre')\n",
    "plt.ylabel('Prediction')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(880,) (880,)\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "       Blues       0.54      0.69      0.61       110\n",
      "   Classical       0.86      0.95      0.90       110\n",
      "     Country       0.57      0.78      0.66       110\n",
      "     Hip-Hop       0.83      0.73      0.78       110\n",
      "        Jazz       0.69      0.75      0.72       110\n",
      "       Metal       0.77      0.80      0.78       110\n",
      "         Pop       0.68      0.35      0.47       110\n",
      "        Rock       0.55      0.40      0.46       110\n",
      "\n",
      "    accuracy                           0.68       880\n",
      "   macro avg       0.69      0.68      0.67       880\n",
      "weighted avg       0.69      0.68      0.67       880\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "\n",
    "print(Y_Test.shape, y_pred.shape)\n",
    "print(classification_report(Y_Test, y_pred, target_names=list(le.classes_)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Blues</th>\n",
       "      <td>0.542857</td>\n",
       "      <td>0.690909</td>\n",
       "      <td>0.608000</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Classical</th>\n",
       "      <td>0.859504</td>\n",
       "      <td>0.945455</td>\n",
       "      <td>0.900433</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <td>0.573333</td>\n",
       "      <td>0.781818</td>\n",
       "      <td>0.661538</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Hip-Hop</th>\n",
       "      <td>0.833333</td>\n",
       "      <td>0.727273</td>\n",
       "      <td>0.776699</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Jazz</th>\n",
       "      <td>0.685950</td>\n",
       "      <td>0.754545</td>\n",
       "      <td>0.718615</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Metal</th>\n",
       "      <td>0.765217</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.782222</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Pop</th>\n",
       "      <td>0.684211</td>\n",
       "      <td>0.354545</td>\n",
       "      <td>0.467066</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Rock</th>\n",
       "      <td>0.550000</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>0.463158</td>\n",
       "      <td>110.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>accuracy</th>\n",
       "      <td>0.681818</td>\n",
       "      <td>0.681818</td>\n",
       "      <td>0.681818</td>\n",
       "      <td>0.681818</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>macro avg</th>\n",
       "      <td>0.686801</td>\n",
       "      <td>0.681818</td>\n",
       "      <td>0.672216</td>\n",
       "      <td>880.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted avg</th>\n",
       "      <td>0.686801</td>\n",
       "      <td>0.681818</td>\n",
       "      <td>0.672216</td>\n",
       "      <td>880.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              precision    recall  f1-score     support\n",
       "Blues          0.542857  0.690909  0.608000  110.000000\n",
       "Classical      0.859504  0.945455  0.900433  110.000000\n",
       "Country        0.573333  0.781818  0.661538  110.000000\n",
       "Hip-Hop        0.833333  0.727273  0.776699  110.000000\n",
       "Jazz           0.685950  0.754545  0.718615  110.000000\n",
       "Metal          0.765217  0.800000  0.782222  110.000000\n",
       "Pop            0.684211  0.354545  0.467066  110.000000\n",
       "Rock           0.550000  0.400000  0.463158  110.000000\n",
       "accuracy       0.681818  0.681818  0.681818    0.681818\n",
       "macro avg      0.686801  0.681818  0.672216  880.000000\n",
       "weighted avg   0.686801  0.681818  0.672216  880.000000"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Create pandas df from classification report\n",
    "report = classification_report(Y_Test, y_pred, target_names=list(le.classes_), output_dict=True)\n",
    "reportdf = pd.DataFrame(report)\n",
    "reportdf.T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Export report to latex\n",
    "reportdf.to_latex('Class_Report_Timbre_' + mtype + '.tex', float_format=\"%.3f\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}