Description:DESCRIPTION:
Field of the invention:
[0001] The present disclosure generally relates to the technical field of diagnosis and therapy of autoimmune diseases and in specific, relates to a system that analyses user medical data and predicts autoimmune diseases with an emphasis on arthritis.
Background of the invention:
[0002] Up to 3% of the general population suffers from autoimmune illnesses, which is a serious health problem. Even though these diseases have different clinical presentations, they have several things in common, such as geographic distributions, population frequencies, therapeutic approaches, and some clinical traits that could be signs of shared pathophysiological pathways. Furthermore, the presence of shared environmental and genetic variables that predispose a person to autoimmunity is supported by the co-occurrence of several autoimmune illnesses in the same person or family.

[0003] The synovial lining of the afflicted joint becomes infiltrated by activated lymphocytes and macrophages, which is a defining feature of the chronic, progressive disease known as RA. These cells generate cytokines and degradative enzymes that cause joint damage and inflammation, frequently resulting in chronic disability. Systemic in nature, RA frequently manifests as extra-articular symptoms that can range in severity from somewhat unimportant issues like rheumatoid nodules to potentially fatal organ dysfunction.

[0004] Clinically, RA is a tremendously diverse illness that can range from moderate to very debilitating, with up to one patient out of twenty developing severe, erosive disease. Early in the course of the illness, there is joint injury, and the first six years are when joint abnormalities advance the fastest. As many as 70% of patients display some radiological signs of joint degeneration within three years of the disease's beginning.

[0005] In existing technology, the arthritis profile data (APD) dataset obtained from a clinical laboratory. These data serves incomplete dataset with missing values. To address this issue imputation methods are employed to convert the incomplete dataset into a reasonably complete one. In addition, the work addresses the challenge of finding the optimal machine learning model for the Arthritis Profile Data to predict disease with high accuracy. However, the existing system does not predict missing data with high accuracy. The system does not provide optimal performance for a specific problem.

[0006] Therefore, there is a need for a system that provides optimal performance for a specific problem. There is also a need for a system that predicts Arthritis disease with high accuracy. There is also a need for a system that improves the classification and prediction of autoimmune diseases and identify the hidden biomarkers in the APD dataset. There is also a need for a system that identifies the hidden biomarkers. There is also a need for a system that contributes to the development of new therapeutic targets and innovative treatments for arthritis.
Objectives of the invention:
[0007] The primary objective of the invention is to provide a system that analyses user medical data and predicts autoimmune diseases with an emphasis on arthritis.

[0008] Another objective of the invention is to provide a system that to assist medical practitioners in early predicting autoimmune diseases.

[0009] The other objective of the invention is to provide a system that analysis the disease, selects suitable computational techniques, improve the algorithm for the predicted model, thereby developing an application model for medical practitioners.

[0010] Yet another objective of the invention is to provide a system that improves the efficiency and effectiveness of the selection process.

[0011] Further objective of the invention is to provide a system that predicts and identifies identify the hidden biomarkers in the APD dataset with high accuracy, thereby providing optimal performance for specific problem.
Summary of the invention:
[0012] The present disclosure proposes a phenomenal approach for autoimmune arthritis disease prediction model using machine learning and metaheuristic algorithms. The following presents a simplified summary in order to provide a basic understanding of some aspects of the claimed subject matter. This summary is not an extensive overview. It is not intended to identify key/critical elements or to delineate the scope of the claimed subject matter. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is presented later.

[0013] In order to overcome the above deficiencies of the prior art, the present disclosure is to solve the technical problem to provide a system that analyses user medical data and predicts autoimmune diseases with an emphasis on arthritis.

[0014] According to an aspect, the invention provides a phenomenal approach for autoimmune arthritis disease prediction model using machine learning and metaheuristic algorithms. The system for predicting autoimmune arthritis disease comprises a computing device having a processor and memory for storing one or more instruction. The processor is configured to execute plurality of modules to perform an operation.

[0015] The processor is in communication with an application server via a network. The plurality of modules comprises a data acquisition module, a noise removal module, an imputation module, an evaluation module, a comparison module, an identification module and a prediction module.

[0016] The data acquisition module is configured to collect data from at least user and calculates missing datasets from the collected data, thereby transmitting the collected data. The noise removal module is configured to receive the collected data from the data acquisition module and remove noise from the received data. The imputation module is configured to replace the missing values using machine learning imputation techniques. In specific, the multiple imputation techniques include mean imputation, median imputation, mode imputation, random value imputation, kNN imputation, MICE imputation and random forest imputation.

[0017] The evaluation module is configured to identify the best imputation techniques for the APD dataset using different imputation methods. In specific, the different imputation methods include Imputed Mean Average Error (MAEim), Imputed Mean Square Error (MSEim), Imputed Root Mean Square Error (RMSEim), and Imputed Coefficient of Determination or R-Squared (R2im). The comparison module is configured to compare different machine learning algorithms and identify algorithm that is suitable for the APD dataset.

[0018] The identification module is configured to identify the metaheuristic algorithms to determine optimal features in the APD dataset. In specific, the metaheuristic algorithms include Particle Swarm Optimization, Grey Wolf Optimization, Cuckoo Search, Genetic Algorithm and Jaya Algorithm are implemented to identify the optimal features in the APD dataset.

[0019] The prediction module is configured to predict autoimmune diseases by identifying best metaheuristic algorithm for the dataset, thereby utilizing the identified features for the prediction. From the comparison module, XGBoost classifier is suitable for the APD dataset prediction. Hyperparameter tuning for the XGBoost classifier is performed using different hyper parameter techniques, which include Bayesian Optimization, GridSearch CV, Halving GridSearch, Randomized Search CV, and Halving Randomized and found that GridSearch CV is best. The system for predicting autoimmune arthritis diseases achieves highest prediction accuracy of 98.7%.

[0020] According to another aspect, the invention provides a method for operating a system for predicting autoimmune arthritis diseases. At one step, a data acquisition module collects data from at least user and calculates missing values from the collected data, thereby transmitting the collected data. At one step, a noise removal module receives the collected data from the data acquisition module and impute the missing values from the received data. At one step, an imputation module replaces the missing values using machine learning imputation techniques.

[0021] At one step, an evaluation module identifies the best imputation techniques for the APD dataset using different imputation methods. In specific, the imputation method is best using degree of proximity and degree of residual concept. At one step, a comparison module compares different machine learning algorithm namely Logistic Regression, K-nearest neighbors, support vector machine, random forest and XGBoost classifier. XGBoost classifier is the best.

[0022] At one step, an identification module identifies the metaheuristic algorithms for determing optimal features in the APD dataset. In specific, the metaheuristic algorithm is suitable using cross entropy in objective function, thereby modifying objective function using machine learning (ML). At one step, a prediction module predicts autoimmune diseases by identifying best metaheuristic algorithm for the dataset, thereby utilizing the identified features for the prediction. In specific, the best metaheuristic algorithm for the dataset and use that identified features for the prediction on XGBoost Classifier.

[0023] Further, objects and advantages of the present invention will be apparent from a study of the following portion of the specification, the claims, and the attached drawings.
Detailed description of drawings:
[0024] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate an embodiment of the invention, and, together with the description, explain the principles of the invention.

[0025] FIG. 1A illustrates a block diagram of a system for predicting autoimmune arthritis diseases, in accordance to an exemplary embodiment of the invention.

[0026] FIG. 1B illustrates a block diagram of plurality of modules in the system for predicting autoimmune arthritis diseases, in accordance to an exemplary embodiment of the invention.

[0027] FIG. 2 illustrates a systematic approach for APD dataset with suitable imputation techniques, in accordance to an example embodiment of the invention.

[0028] FIG. 3 illustrates a systematic approach for APD dataset with suitable Machine Learning model, in accordance to an example embodiment of the invention.

[0029] FIG. 4 illustrates an architecture framework of the APD dataset predictive model, in accordance to an example embodiment of the invention.

[0030] FIG. 5 illustrates a flowchart of a method for operating the system for predicting autoimmune arthritis diseases, in accordance to an exemplary embodiment of the invention.
Detailed invention disclosure:
[0031] Various embodiments of the present invention will be described in reference to the accompanying drawings. Wherever possible, same or similar reference numerals are used in the drawings and the description to refer to the same or like parts or steps.

[0032] The present disclosure has been made with a view towards solving the problem with the prior art described above, and it is an object of the present invention to provide a system that analyses user medical data and predicts autoimmune diseases with an emphasis on arthritis.

[0033] According to an exemplary embodiment of the invention, FIG. 1A refers to a phenomenal approach for autoimmune arthritis disease prediction model using machine learning and metaheuristic algorithms. The system 100 for predicting autoimmune arthritis disease comprises a computing device 102 having a processor 104 and memory 106 for storing one or more instruction. The processor 104 is configured to execute plurality of modules 108 to perform an operation.

[0034] In one embodiment herein, the processor 104 is in communication with an application server 112 via a network 110. The plurality of modules 108 comprises a data acquisition module 114, a noise removal module 116, an imputation module 118, an evaluation module 120, a comparison module 122, an identification module 124 and a prediction module 126 as depicted in FIG. 1B.

[0035] In one embodiment herein, the data acquisition module 114 is configured to collect data from at least user and calculates missing datasets from the collected data, thereby transmitting the collected data. In one embodiment herein, the noise removal module 116 is configured to receive the collected data from the data acquisition module 114 and remove noise from the received data. In one embodiment herein, the imputation module 118 is configured to replace the missing values using machine learning imputation techniques. In specific, the multiple imputation techniques include mean imputation, median imputation, mode imputation, random value imputation, kNN imputation, MICE imputation and random forest imputation.

[0036] In one embodiment herein, the evaluation module 120 is configured to identify the best imputation techniques for the APD dataset using different imputation methods. In specific, the different imputation methods include Imputed Mean Average Error (MAEim), Imputed Mean Square Error (MSEim), Imputed Root Mean Square Error (RMSEim), and Imputed Coefficient of Determination or R-Squared (R2im). In one embodiment herein, the comparison module 122 is configured to compare different machine learning algorithms and identify algorithm that is suitable for the APD dataset.

[0037] In one embodiment herein, the identification module 124 is configured to identify the metaheuristic algorithms to determine optimal features in the APD dataset. In specific, the metaheuristic algorithms include Particle Swarm Optimization, Grey Wolf Optimization, Cuckoo Search, Genetic Algorithm and Jaya Algorithm are implemented to identify the optimal features in the APD dataset.

[0038] In one embodiment herein, the prediction module 126 is configured to predict autoimmune diseases by identifying best metaheuristic algorithm for the dataset, thereby utilizing the identified features for the prediction. From the comparison module 122, XGBoost classifier is suitable for the APD dataset prediction. Hyperparameter tuning for the XGBoost classifier is performed using different hyper parameter techniques, which include Bayesian Optimization, GridSearch CV, Halving GridSearch, Randomized Search CV, and Halving Randomized and found that GridSearch CV is best. The system 100 for predicting autoimmune arthritis diseases achieves highest prediction accuracy of 98.7%.

[0039] According to an exemplary embodiment of the invention, FIG. 2 refers to an architecture framework 200 of the APD dataset predictive model. In one embodiment herein, first, at step 202, a comma separated values (CSV) file is created from the collected data. Later, at step 204, data pre-processing is required as the APD dataset contains missing values. In one embodiment herein, little’s MCAR test is conducted in order to determine a category of the missing data at step 206. The little’s MCAR test guides the implementation of various imputation techniques on the incomplete dataset at step 208.

[0040] Later, at step 210, several imputation techniques including mean imputation, median imputation, mode imputation, random value imputation, kNN imputation, MICE imputation, and random forest imputation are applied to create seven imputed APD datasets. In one embodiment herein, statistical properties such as mean, median, standard deviation, skewness and kurtosis are determined for both the incomplete datasets and the seven complete imputed datasets, thereby evaluating the effectiveness of the imputation techniques at step 212. In one embodiment herein, the statistical properties of the imputed datasets are compared with those of the observed dataset to assess their distribution.

[0041] In one embodiment herein, the degree of proximity is calculated to identify the imputed values that closely resemble the original dataset values for each of the seven imputed datasets. In one embodiment herein, Imputed Mean Average Error (MAEim), Imputed Mean Square Error (MSEim), Imputed Root Mean Square Error (RMSEim) and Imputed Coefficient of Determination or R-Squared (R2im) are computed between each imputed dataset and the original incomplete APD dataset (where missing values are replaced by zero) for each attribute in the dataset.

[0042] At step 214, performance metrics are typically calculated by comparing the actual target values of a dataset with the predicted values obtained from a machine learning algorithm applied to the same dataset. In this case, the degree of residual is determined for MAEim, RMSEim, and R2im to assess the performance of the imputed dataset. In one embodiment herein, the degree of residual quantifies the difference (error) between the original and imputed values. Once these performance metrics are calculated, the first step is to rank each imputed dataset with respect to attributes. Then, the degree of residual procedure is applied for each performance metric. A higher degree of residual indicates a lower residual in the imputed dataset.

[0043] In one embodiment herein, the degree of similarity is computed between each imputed dataset and an equal-sized complete APD dataset (with missing values replaced by zero) and between each imputed dataset and an unequal-sized incomplete APD dataset (ignoring the missing values). Cohen's d is calculated to determine the degree of similarity for equal and unequal sizes of imputed datasets, which evaluates the effect of size. Eventually, the degree of residual and proximity serve as the final factors in determining the most effective imputation techniques for the APD dataset at step 216.

[0044] According to an exemplary embodiment of the invention, FIG. 3 refers to a systematic approach for APD dataset with suitable machine learning model. The proposed methodology involves the analysis of different disease datasets, including APD, breast cancer, cardiovascular disease, diabetes, kidney disease, and RA. The following steps are carried out in the proposed methodology. First, at step 302, benchmark datasets are obtained from the Kaggle repository, including APD from the laboratory and RA from the ACR/EULAR RA classification criteria. Later, at step 304, basic preprocessing steps are applied to each disease dataset to check for the missing values and handle them using imputation techniques. Extraneous features are removed from the datasets. In Python, categorical data is transformed into dummy indicator values using pandas.get_dummies ().

[0045] Later, at step 306, the hold-out method is applied to the APD and benchmark datasets. The dataset is split into training and testing data in different proportions (e.g., 80:20, 70:30, 60:40). Supervised machine learning models such as LR, KNN, and SVM are trained, and techniques such as RF and XGBoost are ensembled on the training datasets to find the best accuracy. Later, at step 308, cross-validation is performed on the APD and benchmark datasets. Cross-validation folds (e.g., 3-fold, 5-fold, 10-fold) are used. The machine learning models are trained, and techniques are ensembled on different folds of the cross-validation datasets to find the best accuracy.

[0046] At step 310, the training recognizes that the size of the test data and the number of cross-validation folds significantly impact the accuracy of the machine learning models. The empirical evidence shows that increasing the test size and the number of folds can lead to varying performance results. In one embodiment herein, some models exhibit increased performance, while others may experience a decrease in accuracy. The XGBoost boosting ensemble technique is deemed suitable for the APD dataset because it handles small-sized and high-dimensional datasets efficiently.

[0047] In one embodiment herein, the proposed methodology aims to analyze and compare the performance of various machine learning algorithms and ensemble techniques on different disease datasets, considering different preprocessing steps, hold-out methods, and cross-validation folds. The goal is to identify the best-performing models and techniques for each disease dataset.

[0048] According to an exemplary embodiment of the invention, FIG. 4 refers to an architecture framework of the APD dataset predictive model. FIG. 2 and FIG. 3 are integrated with Figure 4 to illustrate the task carried out before going to Phase 3 that narrates the different hyperparameter tuning methods involved in the XGBoost classifier and metaheuristic techniques used to improve the XGBoost classifier.

[0049] First, at step 402, the method focuses on studying autoimmune diseases, particularly arthritis, and analyzing the use of medical and biological data for classification and prediction in autoimmune diseases. The conventional model designed suitable imputation techniques and a machine learning model to classify and predict autoimmune diseases using the APD dataset. The proposed method aims to identify hidden biomarkers and the optimal subset of features in the APD dataset to maximize the classification performance of the XGBoost model at step 404.

[0050] In one embodiment herein, using the nature-inspired optimization algorithms with cross-entropy as the objective function, aim to find the hyperparameters that increase the performance of the XGBoost classifier for the APD dataset. Several hyperparameter optimization techniques employed, including Bayesian Optimization, GridSearch CV, Halving GridSearch, Randomized Search CV, and Halving Randomized, and found that GridSearch CV achieved the highest accuracy of 97.1% at step 406.

[0051] At step 408, the method evaluate and compare the performance of metaheuristic algorithms that are efficient and effective for the APD dataset using Cross Entropy in Objective Function for Metaheuristic Algorithms using XGBoost Classifier (CEOMAX). Computational intelligence was applied with nature-inspired techniques such as Particle Swarm Optimization (PSO), Grey Wolf Optimization, Cuckoo Search, Genetic Algorithm, and Jaya Algorithm, using cross entropy as the objective function. Among these algorithms, PSO with cross entropy as the objective function and the XGBoost classifier achieved the highest prediction accuracy of 98.7%.

[0052] At step 410, the method focused on finding the hyperoptimal subset of features in the APD dataset using computational intelligence with nature-inspired techniques, aims to utilize Iterative Mode-Based Hyperoptimal Feature Selection (IMHFS) in CEOMAX. Metaheuristic algorithms have a stochastic nature, which means that they can produce a range of solution qualities across multiple trials. To address this variability, 25 iterations performed and identified the best features by taking the mode of the significant features obtained from these iterations.

[0053] The proposed algorithm CEOMAX alone with IMHFS obtained the hyperoptimal subset of features with an accuracy of 96.78%, outperforming other metaheuristic algorithms, the integration of cross entropy on PSO with XGBoost is the best. The hyperoptimal features identified in this case are CRP, ESRo, ASO, RF, Uric Acid, ESRh, Calcium, and RBC, with a prediction accuracy of 96.78%. Additionally, discovered hidden biomarkers such as ASO, Uric_Acid, Calcium, and RBC, which can enhance the understanding of the underlying mechanisms of arthritis.

[0054] In one embodiment herein, the proposed model aims to improve the classification and prediction of autoimmune diseases, specifically arthritis, by identifying optimal features, hidden biomarkers, and hyperparameters, which can contribute to the development of new therapeutic targets and innovative treatments for arthritis.

[0055] According to another embodiment of the invention, FIG. 5 refers to a flowchart 500 of a method for operating a system for predicting autoimmune arthritis diseases. At step 502, a data acquisition module 114 collects data from at least user and calculates missing datasets from the collected data, thereby transmitting the collected data. At step 504, a noise removal module receives 116 the collected data from the data acquisition module 114 and removing noise from the received data. At step 506, an imputation module 118 replaces the missing values using machine learning imputation techniques.

[0056] At step 508, an evaluation module 120 identifies the best imputation techniques for the APD dataset using different imputation methods. At step 510, a comparison module 122 compares different machine learning algorithms and identify algorithm that is suitable for the APD dataset. At step 512, an identification module 124 identifies the metaheuristic algorithms for determing optimal features in the APD dataset. At step 514, a prediction module 126 identifies best metaheuristic algorithm for the dataset, thereby utilizing the identified features for the prediction. In specific, the best metaheuristic algorithm for the dataset and use that identified features for the prediction on XGBoost Classifier.

[0057] Numerous advantages of the present disclosure may be apparent from the discussion above. In accordance with the present disclosure, a phenomenal approach for autoimmune arthritis disease prediction model using machine learning and metaheuristic algorithms. The proposed system 100 analyses user medical data and predicts autoimmune diseases with an emphasis on arthritis. The proposed system 100 assists medical practitioners in early predicting autoimmune diseases.

[0058] The proposed system 100 analysis the disease, selects suitable computational techniques, improve the algorithm for the predicted model, thereby developing an application model for medical practitioners. The proposed system 100 improves the efficiency and effectiveness of the selection process. The proposed system 100 predicts missing data with high accuracy, thereby providing optimal performance for specific problem.

[0059] It will readily be apparent that numerous modifications and alterations can be made to the processes described in the foregoing examples without departing from the principles underlying the invention, and all such modifications and alterations are intended to be embraced by this application.

, Claims:CLAIMS:
I / We Claim:
1. A system (100) for predicting autoimmune arthritis disease, comprising:
a computing device (102) having a processor (104) and memory (106) for storing one or more instruction, wherein the processor (104) is configured to execute plurality of modules (108) to perform an operation, wherein the plurality of modules (108) comprises:
a data acquisition module (114) configured to collect data from at least user and calculates missing datasets from the collected data, thereby transmitting the collected data;
a noise removal module (116) configured to receive the collected data from the data acquisition module (114) and remove noise from the received data;
an imputation module (118) configured to replace the missing values using machine learning imputation techniques;
an evaluation module (120) configured to identify the best imputation techniques for the APD dataset using different imputation methods;
a comparison module (122) configured to compare different machine learning algorithms and identify algorithm that is suitable for the APD dataset;
an identification module (124) configured to identify the metaheuristic algorithms to determine optimal features in the APD dataset; and
a prediction module (126) configured to identify best metaheuristic algorithm for the dataset, thereby utilizing the identified features for the prediction.
2. The system (100) for predicting autoimmune arthritis disease as claimed in claim 1, wherein the machine learning imputation techniques include mean imputation, median imputation, mode imputation, random value imputation, kNN imputation, MICE imputation and random forest imputation.
3. The system (100) for predicting autoimmune arthritis disease as claimed in claim 1, wherein the different imputation methods include Imputed Mean Average Error (MAEim), Imputed Mean Square Error (MSEim), Imputed Root Mean Square Error (RMSEim), and Imputed Coefficient of Determination or R-Squared (R2im).
4. The system (100) for predicting autoimmune arthritis disease as claimed in claim 1, wherein different hyperparameter tuning user for XGBoost classifier, which include Bayesian Optimization, GridSearch CV, Halving GridSearch, Randomized Search CV, and Halving Randomized and found that GridSearch CV.
5. The system (100) for predicting autoimmune arthritis diseases as claimed in claim 1, wherein the system (100) achieves highest prediction accuracy of 98.7%.
6. The system (100) for predicting autoimmune arthritis diseases as claimed in claim 1, wherein the machine learning (ML) determines the degree of residual to access the performance of the collected data.
7. The system (100) for predicting autoimmune arthritis diseases as claimed in claim 1, wherein the computing device (102) includes a smartphone, a computer, a tablet and a personal digital assistant (PDA).
8. The system (100) for predicting autoimmune arthritis diseases as claimed in claim 1, wherein the processor (104) is in communication with an application server (112) via a network (110).
9. A method for operating a system (100)for predicting autoimmune arthritis diseases, comprising:
collecting, by a data acquisition module (114), data from at least user and calculates missing datasets from the collected data, thereby transmitting the collected data;
receiving, by a noise removal module (116), the collected data from the data acquisition module (114) and removing noise from the received data;
replacing, by an imputation module (118), the missing values using machine learning imputation techniques;
identifying, by an evaluation module (120), the best imputation techniques for the APD dataset using different imputation methods;
comparing, by a comparison module (122), different machine learning algorithms and identifying algorithm that is suitable for the APD dataset;
identifying, by an identification module (124), the metaheuristic algorithms for determing optimal features in the APD dataset; and
identifying, predicting, by a prediction module (126), best metaheuristic algorithm for the dataset, thereby utilizing the identified features for the prediction.