Sleeping Problem Diagnosis

王绪爽 赵育衡

Abstract：As the rate of insomnia in Chinese is experienced a upward trend， people started to pay attention to the sleep problem and human body health. Therefore， we are required to analyze the data about human sleeping.We use MLP to train neural models， and then we have a neural model based on patient information to diagnose patients. In addition， we introduce roc line to check whether the fitting result is accurate. Some roc areas are relatively large， which indicates that machine learning is progressing smoothly.

Key words：multilayer perception；neural network；factor analysis；sleeping problems and diagnosis

中圖分类号：R740 文献标识码：A 文章编号：1671-2064（2018）12-0217-02

1 Introduction

The World Sleep Day draws peoples attention to the importance of sleeping quality. The sleeping quality affects our mental state of a whole day. As the rate of the insomnia in Chinese are in an upward trend， its necessary to do some research or analysis about the sleeping problems.

Two major factors related to the sleeping problems are objective and subjective factors.The former one might be the medicine or drinks we have before sleep and the latter one is our mental pressure or working stress， etc. Especially for young people who is in the period of growth and development are easily have functional disorder caused by the insomnia.

2 Notations（Table 1）

3 Analysis of the Diagnosis

3.1 Classifying the sleep problems

Based on the information， we divided the diagnosis result as 9 types according to their characteristic and pathogenic. Specific divided results are as follows Figure 1：

3.2 Introducing the MLP and its advantages

In this case， we introduce Multilayer Perception（referred as MLP） of the neural network to the question. MLP has a great fit into the question for the following reasons：

MLP is good at fitting which can figure out a relationship between diagnosis and sleep

One of the advantages of the machine learning is bigger data we have， more accurate result we get.

3.3 Results and analysis

We use the SPSS to analyze the relationship. We set the diagnosis as dependent variable， all the sleep ranking as the covariant. Besides， because age and sex is discrete and given literally， we set them as factors. Then we start MLP analyzing. Here is the Neural network structure diagram weve got：

The specific explanation for it is as follow Table 2：

As we can see， there are 11 neuros in the hidden layer Table 3.

Dependent Variable： Diagnosis

Error computations are based on the testing sample.

In general， the percent incorrect predictions of training is 24.3%， training time is short， which indicate that the model has a good performance Table 4.

Dependent Variable： Diagnosis

We introduce the ROC line which is to show the fit accuracy. The area between the colorful curve and the black straight line indicates the accuracy. The larger is the area， the more accurate is the fitting.

From the Table 4， because the data of dependent variables（Y1，Y3，Y6，Y7） is bigger， the accuracy of the prediction is higher. In the same way， their area of them in Fig 2 is larger. But for（Y2，Y4，Y8，Y9）， the numbers of the data is not big enough for a relatively accurate prediction result. Also， the area of them is smaller.

4 ROC line

From the Roc line we can see the goodness of the fit， the area of （Y2，Y3，Y6） is relatively large， indicates the fit goes well.

The neural model that we trained is given below. Through the model， we can use the patients sleep condition to make the diagnosis Table 5.

References

[1]José S. Loredo Sonia Ancoli-Israel Joel E. Dimsdale， Sleep quality and blood pressure dipping in obstructive sleep apnea， AJH ， volume number：page2 to 7， 01 September 2001.

[2]Peter M. A. Calverley， Vlasta Brezinova， Neil J. Douglas， James R.Catterall，and David C. Flenley，The Effect of Oxygenation on Sleep Quality in Chronic Bronchitis and Emphysema，ATSJournals，volume number：page3 to 8，July21，1981.

[3]J. Richard Jennings， PhD Matthew F. Muldoon， MD Martica Hall， PhD Daniel J. Buysse， MD Stephen B. Manuck， PhD，Self-reported Sleep Quality is Associated With the Metabolic Syndrome， ATSJournals， volume number：page1 to 9， 01 February 2007.

[4]DanielFoleya SoniaAncoli-Israelb PatriciaBritzc JamesWalshd，Sleep disturbances and chronic disease in older adults： Results of the 2003 National Sleep Foundation Sleep in America Survey， ScienceDirect， volume number：page1 to 9， May 2004.