With the exponential developments of wireless networking and inexpensive Internet of Things (IoT), a wide range of applications has been designed to attain enhanced services. Due to the limited energy capacity of IoT devices, energy-aware clustering techniques can be highly preferable. At the same time, artificial intelligence (AI) techniques can be applied to perform appropriate disease diagnostic processes. With this motivation, this study designs a novel squirrel search algorithm-based energy-aware clustering with a medical data classification (SSAC-MDC) model in an IoT environment. The goal of the SSAC-MDC technique is to attain maximum energy efficiency and disease diagnosis in the IoT environment. The proposed SSAC-MDC technique involves the design of the squirrel search algorithm-based clustering (SSAC) technique to choose the proper set of cluster heads (CHs) and construct clusters. Besides, the medical data classification process involves three different subprocesses namely pre-processing, autoencoder (AE) based classification, and improved beetle antenna search (IBAS) based parameter tuning. The design of the SSAC technique and IBAS based parameter optimization processes show the novelty of the work. For showcasing the improved performance of the SSAC-MDC technique, a series of experiments were performed and the comparative results highlighted the supremacy of the SSAC-MDC technique over the recent methods.
More recent techniques, like smart devices, mobile applications, wearable devices, blockchain-based electronic medical recording systems home virtual assistants, biosensors, are initiating a modern era in health care service [
Furthermore, medical sensor and IoT devices readings could be effectively employed in detecting a severe disease within a certain time period. The syntax is determined to carry out the detection method in the cloud-centric platform. The three sub-systems are developed for carrying out disease diagnosis procedures [
Verma et al. [
Kumar et al. [
This study designs a novel squirrel search algorithm based on energy aware clustering with medical data classification (SSAC-MDC) model in an IoT environment. The proposed SSAC-MDC technique involves the design of the squirrel search algorithm based clustering (SSAC) technique to choose proper set of cluster heads (CHs) and construct clusters. Moreover, the medical data classification process involves three different subprocesses namely pre-processing, autoencoder (AE) based classification, and improved beetle antenna search (IBAS) based parameter tuning. For showcasing the improved performance of the SSAC-MDC technique, a series of experiments were performed and the results are inspected extensively.
The rest of the paper is organized as follows. Section 2 offers the proposed model, Section 3 provides the result analysis, and Section 4 gives the conclusion.
In this study, an effective SSAC-MDC technique has been presented to accomplish energy efficiency and disease diagnosis in the IoT environment. The proposed SSAC-MDC technique operates on three major phases namely data collection, SSAC based cluster construction, and medical data classification.
User healthcare information is collected by a data acquisition method that allows low-power sensors, alternate medical gadgets, and a seamless combination of smart. These sensors are integrated with, and around the human body to monitor the function. In this approach, a user body sensor is comprised of implanted and wearable sensors. All the sensors have been concatenated with bio-sensors such as Blood pressure (BP), and ECG EEG, etc. For maintaining data security during transmission, a network is secured by Secure Socket Layer (SSL) to provide privacy and security. The timestamp synchronization of dissimilar classes of sensor nodes has been implemented.
The typical SSA upgrades the places of individuals based on the present season [
If the season goes to summer, every individual that glide to
The Levy is a random walk system that step obey Levy distribution and is computed in
Assume that the IoT system of
Let
Using:
While
At this stage, the collected medical data are examined in different ways to determine the existence of the diseases. The medical data classification module involves data preprocessing, AE based classification, and IBSA based parameter tuning.
Initially, the data value was gathered in the UCI dataset afterward pre-processed that needs noise elimination or exchange of missing data. The noiseless data use for efficiently detecting designs linked with heart disease. The median studentized remaining manner was implemented for removing unwanted or noisy data as it examines the connection amongst data from the dataset. This procedure of noise elimination improves the detection procedure of heart disease. The primary step is for examining the data obtainable from the dataset and computing the median to the missing value [
During the classification process, the AE model receives the preprocessed data as input and classifies the data into distinct classes. As a foundation of the DL method, DNNs or so-called multilayer perceptron (MLP), are utilized for representing a complicated model relating data outputs
AE is an unsupervised learning method where the specific architecture of DNN is leveraged to representation learning or dimensionality reduction. Especially, the aim is to optimally copy its input to output with the representation feature by presenting a lower-dimension embedding layer (or named as a code). An AE contains a decoder function
Whereas
For improvising the classifier results of the AE model, the parameter tuning process takes place using IBSA technique. In BAS method, uses the succeeding 2 rules stimulated by the behaviors of beetle search with antennae that include detecting and searching behaviors. Note that the beetle searches arbitrarily to examine an unknown environment.
Step 1: consider that the location of the longicorn beetle in
Step 2: proposed the search behavior of left-and right-hand sides, correspondingly, for imitating the activity of the beetle antennae:
Step 3: location upgrade technique:
Let
To improve the efficiency of the BSA, the IBSA is derived by the use of oppositional based learning (OBL) concepts. Each nature inspired meta-heuristic algorithm selects the first parameter arbitrarily as the candidate solution. Usually, the random population selection method follows a uniform distribution system. As the learning algorithm follows the black box method and does not need any background data it starts enhancing the candidate solution to attain optimal results until predetermined condition isn't attained. The efficacy of this algorithm is confined to amount of time taken by validating all candidate solutions within the searching region to attain nearby optimum solutions as global optimal. To achieve global optimal all near optimum solutions have to be examined. Therefore, finding of nearby optimum solution as an early guess of parameter which isn't suitable inside the searching region.
Moreover, to increase the exploration ability within the searching region, assume opposite case concurrently with early guesses, next the probability of achieving a nearby optimum solution to global optimal rises [
In this section, a comprehensive results analysis of the SSAC-MDC technique take place on benchmark healthcare data. The results are inspected under different instances and IoT devices.
No. of instances | 2000 | 4000 | 6000 | 8000 | 10000 |
---|---|---|---|---|---|
Sensitivity | |||||
KNN model | 0.9260 | 0.8840 | 0.9320 | 0.9240 | 0.9360 |
Naïve Bayes | 0.8790 | 0.8460 | 0.8640 | 0.8860 | 0.8910 |
SVM algorithm | 0.8320 | 0.8240 | 0.8390 | 0.8240 | 0.8420 |
Decision tree | 0.9330 | 0.9230 | 0.9360 | 0.9690 | 0.9600 |
EEPSOC-ANN | 0.9478 | 0.9497 | 0.9563 | 0.9723 | 0.9786 |
SSAC-MDC | 0.9703 | 0.9754 | 0.9833 | 0.9909 | 0.9978 |
Specificity | |||||
KNN model | 0.8420 | 0.8610 | 0.8730 | 0.8830 | 0.8930 |
Naïve Bayes | 0.8340 | 0.8360 | 0.8690 | 0.8210 | 0.8640 |
SVM algorithm | 0.8020 | 0.8210 | 0.8340 | 0.7840 | 0.8430 |
Decision tree | 0.9260 | 0.9120 | 0.9240 | 0.8860 | 0.9040 |
EEPSOC-ANN | 0.9432 | 0.9349 | 0.9487 | 0.9240 | 0.9238 |
SSAC-MDC | 0.9619 | 0.9557 | 0.9698 | 0.9673 | 0.9523 |
Accuracy | |||||
KNN model | 0.8940 | 0.9130 | 0.8760 | 0.8640 | 0.8930 |
Naïve Bayes | 0.7680 | 0.7860 | 0.7780 | 0.8010 | 0.8240 |
SVM algorithm | 0.7340 | 0.7767 | 0.7560 | 0.7840 | 0.8160 |
Decision tree | 0.9160 | 0.9240 | 0.9040 | 0.9320 | 0.9280 |
EEPSOC-ANN | 0.9349 | 0.9430 | 0.9348 | 0.9486 | 0.9420 |
SSAC-MDC | 0.9540 | 0.9641 | 0.9555 | 0.9781 | 0.9811 |
F-Score | |||||
KNN model | 0.9240 | 0.9030 | 0.9240 | 0.9140 | 0.9090 |
Naïve Bayes | 0.8560 | 0.8440 | 0.8720 | 0.8460 | 0.8370 |
SVM algorithm | 0.8120 | 0.8240 | 0.8640 | 0.8040 | 0.8190 |
Decision tree | 0.9760 | 0.9360 | 0.9320 | 0.9270 | 0.9330 |
EEPSOC-ANN | 0.9814 | 0.9530 | 0.9467 | 0.9452 | 0.9521 |
SSAC-MDC | 0.9980 | 0.9809 | 0.9765 | 0.9679 | 0.9700 |
The comparative
The loss results analysis of the SSAC-MDC technique on the test medical data is offered in
Lastly, a TEC analysis of the SSAC-MDC technique with recent methods is provided in
IoT sensors | SSAC-MDC | EEPSOC | ABC | GWO | ACO |
---|---|---|---|---|---|
100 | 37.00 | 46.00 | 58.00 | 61.00 | 67.00 |
200 | 42.00 | 52.00 | 62.00 | 69.00 | 74.00 |
300 | 48.00 | 59.00 | 67.00 | 73.00 | 77.00 |
400 | 53.00 | 64.00 | 72.00 | 76.00 | 81.00 |
500 | 59.00 | 69.00 | 78.00 | 83.00 | 84.00 |
In this study, an effective SSAC-MDC technique has been presented to accomplish energy efficiency and disease diagnosis in the IoT environment. The proposed SSAC-MDC technique operates on three major phases namely data collection, SSAC based cluster construction, and medical data classification. During medical data classification process, the collected medical data are examined in different ways to determine the existence of the diseases. The medical data classification module involves data pre-processing, AE based classification, and IBSA based parameter tuning. For improvising the classifier results of the AE model, the parameter tuning process takes place using IBSA technique. For assessing the improved efficiency of the SSAC-MDC technique, a series of experimental analysis were performed and the comparative results highlighted the supremacy of the SSAC-MDC technique over the recent methods. In future, the classification results can be improvised by the design of outlier removal and feature selection approaches.