Comparison of Neural Network Training Algorithms for Classification of Heart Diseases

Karim, Hossein; Niakan, Sharareh Rostam; Safdari, Reza

doi:10.11591/ijai.v7.i4.pp185-189

Cited by 28 publications

(28 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This shows that feedforward backpropagation of ANN has had a good impact on improving prediction on PLS model. These findings agree with previous research that PLS-ANN results in a better performance than that of PLS [10,21]. Figure 2 shows the RMSEP of PLS and PLS-ANN when different number of hidden neurons was applied with different type of SG preprocessing.…”

Section: Latent Variablesupporting

confidence: 91%

“…To optimize the random initial weights, the network was trained 1000 times to achieve global prediction performance [20]. Levenberg Marquardt (LM) backpropagation algorithm was selected as training algorithm in this study [21]. The training process will stop when either the maximum number of epochs is reached, the goal performance is achieved, the performance of gradient is below minimum gradient value, the momentum update is exceeded, or the failure validation is more than the maximum amount.…”

Section: Artificial Neural Networkmentioning

confidence: 99%

See 1 more Smart Citation

Artificial neural network and partial least square in predicting blood hemoglobin using near-infrared spectrum

Idrus

Chia

2019

IJEECS

View full text Add to dashboard Cite

<p>Predictive models is crucial in near-infrared (NIR) spectroscopic analysis. Partial least square - artificial neural network (PLS-ANN) is a hybrid method that may improve the performance of prediction in NIR spectroscopic analysis. This study investigates the advantage of PLS-ANN over the well-known modelling in spectroscopy analysis that is partial least square (PLS) and artificial neural network (ANN). The results show that ANN that coupled with first order SG derivatives achieved the best prediction with root mean square error of prediction (RMSEP) of 0.3517 gd/L and coefficient of determination ( ) of 0.9849 followed by PLS-ANN with RMSEP of 0.4368 gd/L and of 0.9787, and PLS with RMSEP of 0.4669 gd/L and of 0.9727. This suggests that the spectrum information may unable to be totally represented by the first few latent variables of PLS and a nonlinear model is crucial to model these nonlinear information in NIR spectroscopic analysis.</p>

show abstract

Section: Latent Variablesupporting

confidence: 91%

Section: Artificial Neural Networkmentioning

confidence: 99%

Artificial neural network and partial least square in predicting blood hemoglobin using near-infrared spectrum

Idrus

Chia

2019

IJEECS

View full text Add to dashboard Cite

show abstract

“…Pattern recognition and function estimation are the reasons why neural networks are utilized in data mining [28]. Several other reseachers [29][30][31] have also employed neural network in extracting critical relationship among the data. HNN is one of the most commonly used neural network models.…”

Section: Logic Programming In Hopfield Neural Networkmentioning

confidence: 99%

Logic mining in football matches

Kho¹,

Kasihmuddin²,

Mansor³

et al. 2020

IJEECS

View full text Add to dashboard Cite

Sports results forecast has became increasingly popular among the fans nowadays. It made predicting the outcome of a sport's match, a new and interesting challenge. This paper presented a logic mining technique to model the results (Win Draw / Lose) of the football matches played in English Premier League, Spanish La Liga and France Ligue 1. In this research, a method namely k satisfiability based reverse analysis method (kSATRA) hybridized with Ant Colony Optimization (ACO) was brought forward to obtain the logical relationship among the clubs in these leagues. The logical rule obtained from the football matches was used to categorize the results of future matches. ACO is a population-based and nature-inspired algorithm to decipher several combinatorial optimization problems. kSATRA made use of the advantages of Hopfield Neural Network and k Satisfiability representation. The data set used in this study included the data of 6 clubs from each league, which composed of all league matches from year 2014 to 2018. The effectiveness of kSATRA with ACO in obtaining logical rule in football matches was tested based on root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and CPU time. Results acquired from the computer simulation showed the robustness of kSATRA in exhibiting the performance of the clubs.

show abstract

“…This study can enable researchers to choose the appropriate algorithm in order to improve network lifetime, where choosing an effective clustering method is the first issue that is faced during the construction of the clusters in the WSNs [14]. We have simulated several scenarios based on three measured parameters which are; Variation between clusters size, Standard deviation for Mean Square Error for intra-distances [15], and the ratio between minimum cluster size and maximum cluster size in the network. The remainder of the current study will be ensued by the ensuing sections; Section Two entails the related works.…”

Section: Introductionmentioning

confidence: 99%

Evaluate the performance of K-Means and the fuzzy C-Means algorithms to formation balanced clusters in wireless sensor networks

Hassan¹,

Shah²,

Othman³

et al. 2020

IJECE

View full text Add to dashboard Cite

The clustering approach is considered as a vital method for wireless sensor networks (WSNs) by organizing the sensor nodes into specific clusters. Consequently, saving the energy and prolonging network lifetime which is totally dependent on the sensors battery, that is considered as a major challenge in the WSNs. Classification algorithms such as K-means (KM) and Fuzzy C-means (FCM), which are two of the most used algorithms in literature for this purpose in WSNs. However, according to the nature of random nodes deployment manner, on certain occasions, this situation forces these algorithms to produce unbalanced clusters, which adversely affects the lifetime of the network. Based for our knowledge, there is no study has analyzed the performance of these algorithms in terms clusters construction in WSNs. In this study, we investigate in KM and FCM performance and which of them has better ability to construct balanced clusters, in order to enable the researchers to choose the appropriate algorithm for the purpose of improving network lifespan. In this study, we utilize new parameters to evaluate the performance of clusters formation in multi-scenarios. Simulation result shows that our FCM is more superior than KM by producing balanced clusters with the random distribution manner for sensor nodes.

show abstract

Comparison of Neural Network Training Algorithms for Classification of Heart Diseases

Cited by 28 publications

References 24 publications

Artificial neural network and partial least square in predicting blood hemoglobin using near-infrared spectrum

Artificial neural network and partial least square in predicting blood hemoglobin using near-infrared spectrum

Logic mining in football matches

Evaluate the performance of K-Means and the fuzzy C-Means algorithms to formation balanced clusters in wireless sensor networks

Contact Info

Product

Resources

About