個體化預測模型之應用-以醫療就診時間預測為例

An Application of Individualized Forecasting Model-Taking Medical Consultation Time Forecasting for Example

周碩亮(Shuo-Liang Chou)

指導教授 ： 劉士豪

中原大學/商學院/資訊管理研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000101

摘要

精確的預測可以提高各種系統的效益，為企業或是機構帶來競爭力。資訊科技演進至今有越來越多的技術可運用，如何利用這些技術讓預測更精準是大家努力的目標。
在醫院與診所的診療時間預測上目前已有多種算法能有效幫助病患預估需要等待的時間。但是當等候的人數少時，每個看診病患情況又皆不相同，使用各種不同的算法會有截然不同的表現。
預測算法有很多種，各有不同種特性，在何種情況使用何種算法能得到較精確的預測到目前為止尚未有研究。因此本研究嘗試以五種算法為基礎，根據每位病患的情況套用算法，以期改善在各種不同的算法上會有不同表現的情形。
經過本研究的實驗，實際將取得之門診資料做預估，並用預測值評估方法評估各項結果，證實各項預估算法皆無法在各種情況下有最佳的預測結果。因此本研究建議可建立一個「個體化預測模型」，先將得到之病患資料和門診資料作分析，再挑選出該次所預估之病患合適的預估算法，可使預測結果相較單一預估算法來得精準、穩定。
改善原本只使用單一算法做預測，預測表現不穩定的情況後，將能增加醫院與診所預估每位病患須等候的時間可靠度，使得病患能夠更有效的利用候診時間，也能使醫院與診所安排診療更有效率。

關鍵字

個體化預測模型 ； 創新醫療應用 ； 候診時間 ； 預測 ； 模型建立

並列摘要

Accurate forecasting can improve the efficiency of various systems and bring competitiveness to enterprises or organizations. There are more and more technologies available in the evolution of information technology, and the goal is to use these technologies to make forecasting more accurate.
There are many algorithms that can help patients forecast how long they will have to wait for a hospital or clinic visit. However, when the waiting list is small and each patient is different, different algorithms will perform very differently. The
There are many forecasting algorithms, each with different characteristics, which algorithm can be used in which situation to get a more accurate forecasting. No study has been done. Therefore, this study attempts to apply the algorithm to each patient on the basis of five algorithms in order to improve the performance of the different algorithms. The results of this study are shown in the following table.
After the experiments in this study, the actual outpatient data was used to make forecast and the results were evaluated using the forecasting value evaluation method to confirm the results. No forecasting algorithms can give the best forecast result under all conditions. Therefore, this study proposes the development of an "individualized forecasting model" to analyze the patient data and outpatient data first, and then to develop an individualized forecasting model. The selection of the appropriate forecasting algorithms for the patient in question will result in more accurate and stable forecasting than a single forecasting algorithms.
By improving the unstable forecasting performance of a single algorithm, hospitals and clinics will be able to increase the number of patients who need to be forecasts for each patient. The reliability of the waiting time allows patients to make more efficient use of their waiting time, and allows hospitals and clinics to arrange treatment more efficiently.

並列關鍵字

Individualized Forecasting Model ； Innovative Medical Applications ； Waiting Time ； Forecasting ； Model Building

參考文獻

英文部分

1. Fry, J., “Appointment in General Practice”, Operational Research Quarterly, Vol. 15, pp.224-232,1964.

2. Alpert, J. J., “Broken Appointments”, Pediatrics, Vol. 34, pp.127, 1964.

3. Huang XM, Thompson A, “Setting Up Waiting Time Targets to Satisfy Patient：an Application of Mathematical Programming”, JMA Journal of Mathmetics Applied in Medicine and Biology, Vol. 12, pp.235-247, 1995.

4. Kurata JH, Nofawa AN, Phillips DM, Hoffman S, Werblum MN, “Patient and ProviderSatisfaction with Medical Care”, Journal of Family Practice, Vol. 35, p176-179, 1992.