Tools and Libraries Used<\/h3>\n
The project makes use of the following Python libraries:<\/p>\n
- \n
- NumPy<\/li>\n
- Pandas<\/li>\n
- Matplotlib<\/li>\n
- Seaborn<\/li>\n
- Scikit-Learn<\/li>\n<\/ul>\n
Download Machine Learning Parkinson\u2019s Disease Detection Project<\/h3>\n
Please download the source code of Machine Learning Parkinson\u2019s Disease Detection Project from the following link: Machine Learning Parkinson\u2019s Disease Detection Project Code.<\/strong><\/a><\/p>\n
Steps for Detecting Parkinson\u2019s Disease Using Machine Learning<\/h3>\n
You must complete a set of steps in order to create this Machine Learning Parkinson\u2019s Disease Detector. We are here to walk you through every stage of developing an app.<\/p>\n
1. Importing the required libraries.<\/p>\n
import numpy as np\r\nimport pandas as pd\r\nimport matplotlib.pyplot as plt\r\nimport seaborn as sns\r\nsns.set_style('darkgrid')\r\n<\/pre>\n2. Once the necessary libraries have been imported, the next step would be to read the dataset and display the dataset as well. The first five rows have been displayed in the screenshot shown below.<\/p>\n
parkinson_data = pd.read_csv(\"parkinsons.data\")\r\nparkinson_data.head()`\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"parkinsons](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/parkinsons-disease-output.webp\")
<\/a><\/p>\n3. The dataset contains multiple columns, which can be seen above. Let\u2019s understand what is the meaning of each attribute\/column present in the dataset.<\/p>\n
Attributes:<\/strong><\/p>\n
- \n
- name – ASCII subject name and recording number<\/li>\n
- MDVP:Fo(Hz) – Average vocal fundamental frequency<\/li>\n
- MDVP:Fhi(Hz) – Maximum vocal fundamental frequency<\/li>\n
- MDVP:Flo(Hz) – Minimum vocal fundamental frequency<\/li>\n
- MDVP:Jitter(%), MDVP:Jitter(Abs), MDVP:RAP, MDVP:PPQ, Jitter:DDP – Several measures of variation in fundamental frequency<\/li>\n
- MDVP:Shimmer,MDVP:Shimmer(dB),Shimmer:APQ3,Shimmer:APQ5,MDVP:APQ,Shimmer:DDA – Several measures of variation in amplitude<\/li>\n
- NHR, HNR – Two measures of the ratio of noise to tonal components in the voice<\/li>\n
- status – The health status of the subject (one) – Parkinson’s, (zero) – healthy<\/li>\n
- RPDE, D2 – Two nonlinear dynamical complexity measures<\/li>\n
- DFA – Signal fractal scaling exponent<\/li>\n
- spread1,spread2,PPE – Three nonlinear measures of fundamental frequency variation<\/li>\n<\/ul>\n
4. Now that there is a clear understanding of the various attributes present in the dataset, the next step would be to use exploratory data analysis to gain more insights as well as discover some hidden patterns or trends present in the dataset.<\/p>\n
parkinson_data.info()\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"parkinsons](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/parkinsons-disease-detection-output.webp\")
<\/a><\/p>\n5. Datasets usually contain extreme values, which are usually called outliers. Outliers present in the dataset can be analysed using the below code. There are some outliers. As we can see, some attributes have huge differences in their 75-percentile value and maximum value, which is evident from the below output.<\/p>\n
parkinson_data.describe()\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"data](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/data-outliers.webp\")
<\/a><\/p>\n6. After exploring the dataset and the various features, one can use data visualisation concepts to present the data and the trends present in the data in a visually appealing format. Data Visualization techniques like pie charts, dist-plot etc. can be used for this purpose.<\/p>\n
parkinson_data['status'].value_counts().plot(kind='pie', autopct = \"%1.0f%%\")\r\nplt.title(\"Distribution in the 'status' Column\")\r\nplt.show()\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"data](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/data-visyalizations-1.webp\")
<\/a><\/p>\nparkinson_data.hist(figsize=(15,12));``\r\n<\/pre>\n
![\"parkinsons](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/parkinsons-detection-output.webp\")
<\/a><\/p>\n7. After representing the data visually, it is now time to perform Feature Engineering, to select the required features which will be used to train the Machine Learning model.<\/p>\n
print(\"Original shape of data: \", parkinson_data.shape)\r\nx = parkinson_data.drop(['status','name'], axis=1)\r\nprint(\"Features shape:\", x.shape)\r\ny = parkinson_data.status\r\nprint(\"Target shape: \", y.shape)\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"ml](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/ml-parkinsons-detection-output.webp\")
<\/a><\/p>\n8. The dataset contains attributes which belong to a wide range. It would be convenient to have all the values belonging to a similar range or a similar scale which would eventually help the machine learning model understand the data in a better way. The MinMaxScaler can be used for this purpose, as shown below.<\/p>\n
from sklearn.preprocessing import MinMaxScaler\r\nscaler = MinMaxScaler((-1, 1))\r\n# fits the data to the MinMaxScaler\r\nX = scaler.fit_transform(x)\r\n<\/pre>\n
The primary purpose of utilising MinMaxScaler is to standardise feature scales and bring them to a similar level, leading to enhanced performance of specific machine learning algorithms. This technique is especially beneficial for algorithms that utilise distance calculations, like support vector machines or k-nearest neighbours.<\/p>\n
9. In the next step, the dataset will be split into training data and testing data. The training data will be used to train the model, and the testing data will be used to evaluate the model\u2019s performance.<\/p>\n
from sklearn.model_selection import train_test_split\r\nx_train, x_test, y_train, y_test=train_test_split(X, y, test_size=0.2) \r\n<\/pre>\n
10. Now that the data is split into training data and testing data, various machine learning models can be trained using the training data. We will be using models like Logistic Regression, Decision Trees, Random Forest Classifier etc.<\/p>\n
from sklearn.linear_model import LogisticRegression\r\n# Importing the Logistic Regression and creating an instance of it\r\nclf = LogisticRegression()\r\n#Train Model\r\nclf.fit(x_train, y_train)\r\n\r\n\r\n# Prediction on Test and Train Set \r\npred_logistic_test = clf.predict(x_test)\r\npred_logistic_train = clf.predict(x_train)\r\n<\/pre>\n
11. Once the model is trained, the model\u2019s performance can be tested using various metrics like confusion matrix, classification report, accuracy score etc.<\/p>\n
from sklearn.metrics import confusion_matrix,accuracy_score, classification_report\r\nprint(\"Training Accuracy: \", accuracy_score(y_train, pred_logistic_train))\r\nprint(\"Test Accuracy: \", accuracy_score(y_test, pred_logistic_test))\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"accuracy](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/accuracy-test.webp\")
<\/a><\/p>\ncm = confusion_matrix(y_test, pred_logistic_test)\r\nax= plt.subplot()\r\nsns.heatmap(cm, annot=True, ax = ax);\r\nax.set_xlabel('Predicted labels')\r\nax.set_ylabel('True labels') \r\nax.set_title('Confusion Matrix')\r\nplt.show()\r\n<\/pre>\nOutput:<\/strong><\/p>\n
![\"parkinsons](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/parkinsons-disease-project-output-2.webp\")
<\/a><\/p>\nprint(\"\\nClassification Report:\")\r\nprint(classification_report(y_test, pred_logistic_test))\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"classification](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/classification-report-2.webp\")
<\/a><\/p>\n12. The next model, which we will be using is the Decision Tree Classifier.<\/p>\n
from sklearn.tree import DecisionTreeClassifier\r\ndt = DecisionTreeClassifier()\r\n# Train model \r\ndt.fit(x_train, y_train)\r\n\r\n\r\npred_dt_test = dt.predict(x_test)\r\npred_dt_train = dt.predict(x_train)\r\n# Making predictions using the data\r\n\r\n\r\nprint(\"Training Accuracy: \", accuracy_score(y_train, pred_dt_train))\r\nprint(\"Test Accuracy: \", accuracy_score(y_test, pred_dt_test))\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"decision](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/decision-tree-classifier.webp\")
<\/a><\/p>\ncm = confusion_matrix(y_test, pred_dt_test)\r\nax= plt.subplot()\r\nsns.heatmap(cm, annot=True, ax = ax);\r\nax.set_xlabel('Predicted labels')\r\nax.set_ylabel('True labels') \r\nax.set_title('Confusion Matrix')\r\nplt.show()\r\n<\/pre>\nOutput:<\/strong>
![\"ml](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/ml-detecting-parkinsos-output.webp\")
<\/a><\/p>\nprint(\"\\nClassification Report:\")\r\nprint(classification_report(y_test, pred_dt_test))\r\n<\/pre>\n
![\"test](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/test-prediction.webp\")
<\/a><\/p>\n13. After the Decision Tree Classifier, the Random Forest Classifier can be used to predict whether a person will have Parkinson\u2019s disease or not.<\/p>\n
from sklearn.ensemble import RandomForestClassifier\r\nrf = RandomForestClassifier()\r\nrf.fit(x_train, y_train)\r\n\r\n\r\ntrain_pred_rf = rf.predict(x_train)\r\npred_rf = rf.predict(x_test)\r\n\r\n\r\nprint(\"Training Accuracy: \",accuracy_score(y_train, train_pred_rf))\r\nprint(\"Test Accuracy: \",accuracy_score(y_test, pred_rf))\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"random](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/random-forest-classifier.webp\")
<\/a><\/p>\ncm = confusion_matrix(y_test, pred_rf)\r\nax= plt.subplot()\r\nsns.heatmap(cm, annot=True, ax = ax);\r\nax.set_xlabel('Predicted labels')\r\nax.set_ylabel('True labels') \r\nax.set_title('Confusion Matrix')\r\nplt.show()\r\n<\/pre>\nOutput:<\/strong><\/p>\n
![\"detecting](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/detecting-parkinsons-output.webp\")
<\/a><\/p>\nprint(\"\\nClassification Report:\")\r\nprint(classification_report(y_test, pred_rf))\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"predict](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/predict-parkinsons-disease-or-not-3.webp\")
<\/a><\/p>\n14. The last algorithm which will be used is the Gradient Boosting Algorithm.<\/p>\n
from sklearn.ensemble import GradientBoostingClassifier\r\ngb = GradientBoostingClassifier()\r\n# Training model\r\ngb.fit(x_train, y_train)\r\n\r\n# Prediction on test and train set\r\ngb_pred_train = gb.predict(x_train)\r\ngb_pred = gb.predict(x_test)\r\n\r\n\r\nprint(\"Training Accuracy: \",accuracy_score(y_train, gb_pred_train))\r\nprint(\"Test Accuracy: \", accuracy_score(y_test, gb_pred))\r\n\r\n<\/pre>\n
Output:<\/strong><\/p>\n
![\"gradient](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/gradient-boosting-algorithm.webp\")
<\/a><\/p>\ncm = confusion_matrix(y_test, gb_pred)\r\nax= plt.subplot()\r\nsns.heatmap(cm, annot=True, ax = ax);\r\nax.set_xlabel('Predicted labels')\r\nax.set_ylabel('True labels') \r\nax.set_title('Confusion Matrix')\r\nplt.show()<\/pre>\nOutput:<\/strong><\/p>\n
![\"machine](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/machine-laerning-parkinsons-disease-output-1.webp\")
<\/a><\/p>\n![\"last](\"https:\/\/techvidvan.com\/tutorials\/wp-content\/uploads\/sites\/2\/2023\/07\/last-algorithm-1.webp\")
<\/a><\/p>\n