| LEADER |
03294nam a22002297a 4500 |
| 001 |
1621679 |
| 041 |
|
|
|a eng
|
| 044 |
|
|
|b مصر
|
| 100 |
|
|
|9 469475
|a Ahmed, Kholoud Ahmed Maher Hamed
|e Author
|
| 245 |
|
|
|a Comparison of Logistic Regression and Artificial Neural Network Models in Determining the Possible Risk Factors of Autism
|
| 260 |
|
|
|b جامعة عين شمس - كلية التجارة
|c 2016
|g أكتوبر
|
| 300 |
|
|
|a 63 - 93
|
| 336 |
|
|
|a بحوث ومقالات
|b Article
|
| 520 |
|
|
|b There is an agreement among all professionals that autism is one of the most puzzling diseases. The increase of autism prevalence cannot be fully stated by advances in diagnostics or sudden genetic shifts but there is a growing agreement among scientists and clinicians that autism ensue from an interaction between environmental and genetic factors. This study will draw a comparison between the results obtained on a given set of data gathered on a sample of Egyptian autistic children with age and sex matched healthy controlled children by using non parametric statistical technique which is Artificial neural network and a parametric technique which is Forward Stepwise logistic regression (LR) to identify the key individual covariates and factors converging on autism occurrence as an outcome/ criterion variable, in an attempt to identify the possible risk factors that can lead to autism and to know which one of those statistical techniques provides higher accuracy in classification of cases. In the present study both techniques show promising results for diagnostic accuracy of classifying the cases witnessed by the area under the ROC curve where the area under the ROC curve AZ(±SD) of the logistic regression analysis was (0.830 ± 0.034) with 95% confidence limit (0.763- 0.897) and the area under the ROC curve of the artificial neural network analysis was (0.872 ± 0.031) with 95% confidence limit (0.812- 0.933). By using the same partition variable in both techniques, the training sample output of the logistic regression yielded sensitivity of 75.90%, specificity of 78.80% and overall accuracy of the model was 77.40% whereas the hold out sample yielded overall accuracy of 76.50%. The training sample output of the Artificial Neural Network yielded sensitivity, specificity and accuracy of 88.90%, 73.10% and 81.10% respectively while the holdout sample yielded overall accuracy of 76.50%. This indicates that the two analytical methods seem to quite fit well with the data and Neural Networks provides higher sensitivity, accuracy and area under ROC curve than logistic regression so we can depend on to classify subjects on autism.
|
| 653 |
|
|
|a إدارة المخاطر
|a معالجة بيانات
|a الإحصاء الرياضي
|
| 692 |
|
|
|b Autism
|b Artificial Neural Network
|b Logistic Regression
|b ROC Curve
|b Sensitivity
|b Specificity
|
| 700 |
|
|
|9 469476
|a Abdel Aal, Medhat Mohamed Ahmed
|e Co-Author
|
| 773 |
|
|
|4 الاقتصاد
|4 الإدارة
|6 Economics
|6 Management
|c 068
|e Scientific Journal for Economic & Commerce
|f Al-Maġallah Al-ʿilmiyyah Lil-Iqtiṣād Wal Tiğārah
|l 004
|m ع4
|o 0527
|s المجلة العلمية للاقتصاد والتجارة
|v 046
|x 2636-2562
|
| 856 |
|
|
|u 0527-046-004-068.pdf
|
| 930 |
|
|
|d y
|p y
|q n
|
| 995 |
|
|
|a EcoLink
|
| 999 |
|
|
|c 871120
|d 871120
|
