Description:Title:
A Framework to recommend Chronic Kidney Disease Prediction models using Multi-Criteria Decision Making

Specification
The invention and the process by which it is to be carried out are specifically described in the following specification.

Field of Invention
The invention relates to the field of computer science, where multi-criteria decision making is crucial for recommending machine learning methods for the early diagnosis of chronic kidney disease.
Background
Chronic kidney disease (CKD), which is currently present in 10% of the world's population and which eventually leads to death, is incredibly painful. It is rapidly spreading and rising to become one of the leading causes of mortality globally. There are 850 million people worldwide who are likely to have kidney disease due to various factors including smoking, adopting poor dietary habits, and not getting enough sleep. Because the kidneys are slowly destroyed and gradually lose their ability to filter the waste from the blood, it is difficult to notice any distinct symptoms in CKD patients in the early stages. It can develop into kidney failure and cardiac disorders if untreated.
Early CKD detection can lower the risk of CKD progression to later stages as well as the occurrence of additional complications, such as cardiovascular disease, bone disorders, vision loss, etc.
Machine learning techniques are now frequently used in healthcare to develop disease prediction models due to the availability of biomedical data. However, the performance of different MLTs for CKD prediction may vary for different accuracy measures. As a result, the selection of the most suitable machine learning technique among various available machine learning techniques for CKD prediction is a challenging task.
This invention proposes an MCDM-based method to evaluate the MLTs for CKD prediction considering various performance measures taken into account altogether. MCDM is one of the most well-known subfields of decision-making. MCDM finds the most suitable choice among the many possibilities accessible, taking into account multiple conflicting criteria.
Objects of the invention
The recommendation of the machine learning techniques for the prediction of chronic kidney disease is the objective of the current disclosure.
In proposed invention, the following objectives are defined:
For the evaluation of machine learning techniques more than one performance measures are considered.
Performance measures are considered simultaneously.

Diagrams and figures:
Brief Description of drawing:
Fig. 1 describes the process of generating a ranking index for CKD prediction models using the MCDM-based approach.
Fig. 2 describes the experimental procedure of the proposed invention.
Description of Drawings:
Fig. 1 describe the framework for recommending the most suitable machine learning technique for chronic kidney disease prediction. Detailed steps are as follows:
Step 1: Apply different available machine learning techniques (MLTs), let's say m, over the chronic kidney disease dataset.
Step 2: Measure the performance of MLTs in terms of the various performance measures (PMs), let's say n.

Step 3: Construct the decision matrix Dm×n , where each entry dij represents the predictive capability of chronic kidney disease detection for the ith machine learning technique on performance measure j.
Step 4: Apply the CoCoSo method to the decision matrix Dm×n. The CoCoSo method is an MCDM method which is based on the integration of simple additive weighting and exponentially weighted product model.
Step 5: Arrange the alternatives (in this study MLTs for predicting chronic kidney disease) in decreasing order of the assessment value obtained by applying CoCoSo.
Step 6: Recommend the MLTs for the prediction of chronic kidney disease that has the highest assessment value.

Fig. 2 describes the validation of the proposed invention using a chronic kidney disease dataset.
Phase 1 (construction of decision matrix):
Build chronic kidney disease prediction models by applying twelve machine learning techniques on CKD dataset. Next, the performance of twelve chronic kidney disease prediction models is measured in terms of six performance measures (Accuracy (A), Recall (R), Specificity (S), False Negative Rate (FNR), Precision (P), and G-measure). The results are stored in a 12 × 6 matrix.

Phase 2 (Apply CoCoSo to recommend MLTs for chronic kidney disease prediction):
Twelve chronic kidney disease prediction models are evaluated using CoCoSo (MCDM method). First, the 12 × 6 matrix obtained from Phase 1 is used as the decision matrix for applying MCDM method- CoCoSo. Next, calculate the overall assessment score of each machine learning technique for chronic kidney disease prediction. Finally, recommend the machine learning technique for chronic kidney disease prediction that gets the highest assessment score.

Advantages of the Invention:
Advantages of the present invention are as follows:
To select the most suitable machine learning technique from various available machine learning techniques for chronic kidney disease prediction in the presence of more than one performance measure.
Optimize the performance of chronic kidney disease prediction models with respect to various performance measures.
The proposed framework in this invention may be used in various domain, like may be used in the health care sector for predicting various other diseases, in software industries to make decisions considering various performance measures taken into account altogether etc.

We Claim:

1. A system and method for recommending the most suitable machine learning technique for chronic kidney disease prediction considering various performance measures simultaneously.

2. In system of claim 1, the idea is to create a novel framework for recommending the most suitable machine learning technique for chronic kidney disease prediction considering various performance measures simultaneously.
3. The system of claim 2, carries following methods to be used under claims:

• Selecting machine learning techniques as inputs.
• Take the chronic kidney disease dataset.
• Data cleaning including various tasks like dealing with missing values, dealing with outliers, handling class imbalance etc.
• Use an MCDM method

4. Recommend the most suitable machine learning technique for chronic kidney disease prediction.

Abstract
Chronic Kidney Disease (CKD) is currently one of the most prevalent disorders affecting public health. Early disease detection may reduce the severity of its effects. Due to machine learning's ability to identify patterns in data, its use in the medical industry has increased. Many CKD prediction models have been developed in past studies utilizing machine learning techniques (MLTs). The performance of MLTs on CKD prediction may vary for different accuracy measures. Thus, the choice of the appropriate machine-learning technique for CKD prediction is a challenging task. This invention presents multi-criteria decision-making (MCDM)-based framework to recommend the most suitable MLT for CKD prediction considering various performance measures taken into account altogether. For examine the proposed framework, an experimental study was conducted to evaluate the performance of twelve machine learning techniques for predicting chronic kidney disease over the CKD dataset of 400 people considering six performance measures taken into account altogether. Results show that the proposed framework can be used as an efficient tool for recommending the most suitable chronic kidney disease prediction considering various performance measures simultaneously.
, C , C , Claims:We Claim:

1. A system and method for recommending the most suitable machine learning technique for chronic kidney disease prediction considering various performance measures simultaneously.

2. In system of claim 1, the idea is to create a novel framework for recommending the most suitable machine learning technique for chronic kidney disease prediction considering various performance measures simultaneously.
3. The system of claim 2, carries following methods to be used under claims:

• Selecting machine learning techniques as inputs.
• Take the chronic kidney disease dataset.
• Data cleaning including various tasks like dealing with missing values, dealing with outliers, handling class imbalance etc.
• Use an MCDM method

4. Recommend the most suitable machine learning technique for chronic kidney disease prediction.