Azman et al 2021a - Revision history

Firioza at 01:38, 13 March 2024

2024-03-13T01:38:01Z

Rimni: /* 2. Literature review */

2022-03-22T13:15:13Z

‎2. Literature review

Rimni at 13:12, 22 March 2022

2022-03-22T13:12:47Z

Rimni at 13:06, 22 March 2022

2022-03-22T13:06:45Z

Rimni at 12:58, 22 March 2022

2022-03-22T12:58:20Z

Show changes

Rimni at 12:26, 18 March 2022

2022-03-18T12:26:12Z

Rimni at 12:25, 18 March 2022

2022-03-18T12:25:24Z

Rimni at 12:24, 18 March 2022

2022-03-18T12:24:15Z

Rimni at 12:22, 18 March 2022

2022-03-18T12:22:59Z

Rimni: /* 3.3 Model development */

2022-03-18T12:21:43Z

‎3.3 Model development

@@ Line 535: / Line 535: @@
 ==Acknowledgements==
-Part of this work was supported by the grant FRGS/1/2020/STG06/UKM/03/1.
+The research was funded by the Ministry of Higher Education(MOHE) trough Fundamental Research Grants Scheme under the grant number FRGS/1/2020/STG06/UKM/03/1.
 ==References==

@@ Line 53: / Line 53: @@
 A number of researchers have attempted to predict breast-cancer recurrence in some sort of pattern, and very few manage to get a high-accuracy model with a small error. Hussain et al. [30] emphasized that a suitable algorithm and architecture of a model are the key to a critical task in prediction. There may be the existence of redundant or highly correlated attributes in data with high dimension that can severely degrade classification accuracy. This poses a compulsive challenge to machine learning. Hence, Howley et al. [31] investigated the use of PCA to resolve the issue. Skittides and Früh [32] acknowledged PCA to reduce the original datasets by applying orthogonal transformation and aborted extra attributes.
-PCA is a classical statistical method that transforms the original dataset into a new uncorrelated attribute dataset called principal component (PC). Classifiers and PCA are applied to a dataset to improve its accuracy. The performance of numbers of supervised classifiers using PCA has been monitored by [23]. Five different classifiers such as SVM, Decision Tree (DT), Logistic Regression (LR), Naïve Bayes (Gaussian NB), and K-Nearest Neighbor (K-NN) were used. For each sample with different classifiers, PCA functionality for dimensionality reduction was applied in the preprocessing, and a confusion matrix was used for performance evaluation. The results indicate performance variations according to each classifier in which Naïve Bayes with sigmoid PCA has the highest accuracy of 99.2%, whereas SVM with a sigmoid function had the worst performance with 83.5% accuracy. With this breast-cancer dataset, KNN performs best among other five supervised machine learning algorithms. A similar study by [33], a prognostic study, detected breast cancer using PCA that had been coupled with a classifier, ANN. The result indicated that applying both PCA and a classifier can lead to an increase in performance. Machine learning is rapidly advancing through deep learning and continues to boost the interest in applying these techniques to improve the accuracy of cancer diagnosis.
+PCA is a classical statistical method that transforms the original dataset into a new uncorrelated attribute dataset called principal component (PC). Classifiers and PCA are applied to a dataset to improve its accuracy. The performance of numbers of supervised classifiers using PCA has been monitored by Mushtaq and  Yaqub [23]. Five different classifiers such as SVM, Decision Tree (DT), Logistic Regression (LR), Naïve Bayes (Gaussian NB), and K-Nearest Neighbor (K-NN) were used. For each sample with different classifiers, PCA functionality for dimensionality reduction was applied in the preprocessing, and a confusion matrix was used for performance evaluation. The results indicate performance variations according to each classifier in which Naïve Bayes with sigmoid PCA has the highest accuracy of 99.2%, whereas SVM with a sigmoid function had the worst performance with 83.5% accuracy. With this breast-cancer dataset, KNN performs best among other five supervised machine learning algorithms. A similar study by Jhajharia et al. [33], a prognostic study, detected breast cancer using PCA that had been coupled with a classifier, ANN. The result indicated that applying both PCA and a classifier can lead to an increase in performance. Machine learning is rapidly advancing through deep learning and continues to boost the interest in applying these techniques to improve the accuracy of cancer diagnosis.
 A DNN has ANN characteristics, with a more complex network structure that has several hidden layers aiming to set new records in accuracy, specifically for image recognition and sound recognition [34]. These hidden layers are used for learning different attribute maps for a diagnosis. Liang et al. [35] explained that medical data are more complex nowadays, and traditional models, for instance, LR and random forest (RF), are thus not efficient in forecasting medical diagnosis. DNN is a powerful pixel classifier and is listed as supervised learning [36]. Many discussions and experiments have been made regarding this issue and the results were observed. Deep learning has been proved to bring a revolution in machine learning, showing improved performance over traditional approaches.

@@ Line 31: / Line 31: @@
 ==1. Introduction==
-World Health Organization announced breast cancer as the most common form of cancer in 2020 and that it has overtaken lung cancer based on statistics by the International Agency for Research on Cancer. According to [1], breast cancer has surpassed lung cancer, with 11.7% new cases in 2020 worldwide for both sexes of all ages. Breast cancer represents one in four common cancers diagnosed among women globally and is the leading cause of cancer death. The American Cancer Society mentioned in [2] that approximately 1 in 8 women will be diagnosed with invasive breast cancer in their lifespan and that 1 in 39 women will die from breast cancer. The incidence and mortality rate in female breast cancer in South-Eastern Asia are 41.2% and 15.0%, respectively.
+World Health Organization announced breast cancer as the most common form of cancer in 2020 and that it has overtaken lung cancer based on statistics by the International Agency for Research on Cancer. According to Global Cancer Observatory [1], breast cancer has surpassed lung cancer, with 11.7% new cases in 2020 worldwide for both sexes of all ages. Breast cancer represents one in four common cancers diagnosed among women globally and is the leading cause of cancer death. The American Cancer Society mentioned in [2] that approximately 1 in 8 women will be diagnosed with invasive breast cancer in their lifespan and that 1 in 39 women will die from breast cancer. The incidence and mortality rate in female breast cancer in South-Eastern Asia are 41.2% and 15.0%, respectively.
 Breast cancer has been recorded as the highest occurrence of all cancers in women worldwide [3-5]. Feng et al. [6] define breast cancer as a compilation of distinct cancerous tumors called malignancies that present in the breast region. Breast cancer is an abnormal and rapid growth of the tissue cells in the breast region. Sainsbury et al. [7] stated that breast cancers are derived from the epithelial cells that line the terminal duct lobular unit. Sariego [8] specified that breast cancer mainly involves mammary glands, which are glands and ducts that deliver the milk. Breast cancer can be categorized into two groups: non-invasive and invasive cancer. A study by Becker [9] demonstrated that the most significant factors for breast cancer are age and low parity. The prevalence of breast cancer is staged according to the size of a tumor that requires different combination of treatments. Surgery, chemotherapy, radiotherapy, and hormone therapy are few common treatments that have been utilized in clinical practice for many years.

@@ Line 124: / Line 124: @@
 |  style="text-align: center;vertical-align: top;"|Negative (0)
 |-style="text-align:center"
 |  style="text-align: center;vertical-align: top;"|Positive (1)
 |  style="text-align: center;vertical-align: top;"|True Positive (TP)

@@ Line 124: / Line 124: @@
 |  style="text-align: center;vertical-align: top;"|Negative (0)
 |-style="text-align:center"
 |  style="text-align: center;vertical-align: top;"|Positive (1)
 |  style="text-align: center;vertical-align: top;"|True Positive (TP)
 |  style="text-align: center;vertical-align: top;"|False Positive (FP)
 |-style="text-align:center"
 |  style="text-align: center;vertical-align: top;"|Negative (0)
 |  style="text-align: center;vertical-align: top;"|False Negative (FN)

@@ Line 117: / Line 117: @@
 {| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;"
 |-style="text-align:center"
-! rowspan='3'style="text-align: center;vertical-align: bottom;"| Predicted Value !! colspan='3' |Actual Value
+! rowspan='4'style="text-align: center;vertical-align: bottom;"| Predicted Value !! colspan='3' |Actual Value
 |-style="text-align:center"
 |  style="text-align: center;vertical-align: top;"|