PREDICTION OF CHRONIC KIDNEY DISEASE USING DEEP NEURAL NETWORK
Abstract : Deep neural Network (DNN) is becoming a focal point in Machine Learning research. Its application is penetrating into different fields and solving intricate and complex problems. DNN is now been applied in health image processing to detect various ailment such as cancer and diabetes. Another disease that is causing threat to our health is the kidney disease. This disease is becoming prevalent due to substances and elements we intake. Death is imminent and inevitable within few days without at least one functioning kidney. Ignoring the kidney malfunction can cause chronic kidney disease leading to death. Frequently, Chronic Kidney Disease (CKD) and its symptoms are mild and gradual, often go unnoticed for years only to be realized lately. Bade, a Local Government of Yobe state in Nigeria has been a center of attention by medical practitioners due to the prevalence of CKD. Unfortunately, a technical approach in culminating the disease is yet to be attained. We obtained a record of 400 patients with 10 attributes as our dataset from Bade General Hospital. We used DNN model to predict the absence or presence of CKD in the patients. The model produced an accuracy of 98%. Furthermore, we identified and highlighted the Features importance to provide the ranking of the features used in the prediction of the CKD. The outcome revealed that two attributes; Creatinine and Bicarbonate have the highest influence on the CKD prediction.
• We found that there is another way to measure CKD as described in the work. Features of age, gender, the existence of diabetes, hypertension, anemia, and cardiovascular disease are used to measure the risk score of CKD using a simple grid search method.
• The existing method treats the existence of related diseases, like diabetes or hypertension, as binary variables, 0/1, to predict CKD risk.
• A large number of features exist in the raw data in which some may cause low information and error; hence feature selection techniques can be used to retrieve useful subset of features and to improve the computation performance.
• The proposed MDFS model performs superior to existing works in terms of accuracy and the number of reduced features.
• The effect of sodium intake on the progression of CKD is controversial, and multiple inconclusive studies exist. No specific recommendation is advisable at this time beyond general recommendations to limit sodium in patients with hypertension and/or fluid overload,
• Currently, no age-specific definitions exist for CKD. In general, older individuals with reduced kidney function are at higher risk for acute kidney injury from pre-renal, renal and post-renal causes.
• The effect of sodium intake on the progression of CKD is controversial, and multiple inconclusive studies exist.
• In any machine learning problem preprocessing of the data before applying any algorithm. Data mining statistical techniques are used for data preprocessing.
• Wrapper methods are better than filter methods in selecting the most useful features but they need a large dataset otherwise there is a problem of overfitting.
• SMOT technique is used to oversample minority class in the dataset in classification problem. In our dataset, we have more patient records having CKD disease than nonCKD. Classifier's performance reduces because of imbalanced classes.
• We used undersampling method and proposed a new cost-sensitive mean-squared error (MSE) loss function to deal with the problem.
• As the creatinine value, which is the target variable, is extremely unbalanced, we used an undersampling method and proposed a cost-sensitive mean squared error (MSE) loss function to deal with the problem.
• Using commonly available health parameters, the proposed system can assess the risk of CKD for public health.
• The proposed model will fail to classify, and the classification result will not be acceptable when we consider actual data.
• They proposed to compare Decision Tree to Support Vector Machine (SVM) technique in predicting CKD. The classification stage revealed that Decision Tree has more successful than SVM recognition in correct classification.
• In this paper, a novel Mixed Data Feature Selection (MDFS) model is proposed to select and filter preeminent features from the medical dataset for earlier CKD prediction, where CKD clinical data with 12 numerical and 12 nominal features are fed to the MDFS model.
• The proposed system aims to identify the most dominant attributes that can aid in the early CKD prediction by involving separate feature selection algorithms on numerical and nominal data.
• It enhances the comprehensibility of data, facilitates better visualization of data, reduces training time of learning algorithms, and improves the performance of prediction.
• They usually work according to a specific learning algorithm which helps in optimizing the performance of a learning algorithm.
• The performance of Decision tree method is 91% accurate when compared with naive Bayes method.
• Various performance parameters like accuracy, specificity, precision, sensitivity, false-positive rate, Fmeasure are used in health care to analyze the performance of the model.
• Even with a small amount of data, a hybrid neural network shows better performance compared to SVM and Random forest.
• Bidirectional Long Short Term memory (BiLSTM) for text analysis and autoencoder network for properties are used. The hybrid neural network is more efficient than statistical methods.
• The advantage of this method is that it can not only increase the importance of rare data, but also avoids the negative effects from high-order exponential errors applied to all samples in the training data set.
|