北京生物醫(yī)學(xué)工程

基于多特征集成決策樹算法的門診需求預(yù)測(cè)

Outpatient demand forecasting based on multi-feature ensemble decision tree algorithm

作者：彭俊張肖建徐超謝勇項(xiàng)薇何達(dá)

單位：寧波大學(xué) 機(jī)械工程與力學(xué)學(xué)院（浙江寧波315211）寧波大學(xué)先進(jìn)儲(chǔ)能技術(shù)與裝備研究院（浙江寧波315211）寧波市鄞州區(qū)婦幼保健院（浙江寧波315211）

關(guān)鍵詞：門診需求預(yù)測(cè); 多特征; GBDT; 隨機(jī)森林; ARIMA

分類號(hào)：R318.04

出版年·卷·期（頁(yè)碼）：2021·40·1（68-73）

摘要：

目的為了準(zhǔn)確預(yù)測(cè)醫(yī)療門診需求量，以便醫(yī)院管理者科學(xué)分配關(guān)鍵醫(yī)療資源，提高服務(wù)效率，本文提出一種基于多特征集成決策樹的醫(yī)療需求預(yù)測(cè)模型。方法首先引入了機(jī)器學(xué)習(xí)算法中的梯度提升決策樹（gradient boosting decision tree, GBDT）和隨機(jī)森林（random forest, RF）。考慮外部因素對(duì)門診人數(shù)的影響，根據(jù)寧波市某婦幼保健院的日產(chǎn)前檢查人數(shù)的歷史數(shù)據(jù)，引入前一天產(chǎn)前檢查人數(shù)、時(shí)間、節(jié)假日、天氣等特征，建立多特征日檢查人數(shù)預(yù)測(cè)模型。預(yù)測(cè)結(jié)果與經(jīng)典自回歸移動(dòng)平均模型(autoregressive integrated moving average model, ARIMA)模型進(jìn)行對(duì)比。結(jié)果 GBDT、RF和ARIMA模型預(yù)測(cè)結(jié)果的平均絕對(duì)百分比誤差（mean absolute percentage error, MAPE）分別是14.95%、17.16%、18.53%。結(jié)論集成決策樹模型在醫(yī)療需求預(yù)測(cè)中的有效性和可行性，并且預(yù)測(cè)精度較傳統(tǒng)的ARIMA模型高。

Objective To predict accurately the demand for medical outpatients, a medical demand forecasting model based on multi-feature ensemble decision tree is proposed. Thus, hospital administrators can scientifically allocate key medical resources and improve service efficiency. Methods Gradient boosting decision tree and random forest (GBDT and RF) in machine learning algorithms were introduced. We considered the influence of external factors on the number of outpatients,according to the historical data of the number of daily prenatal check-ups in a maternal and child health care hospital in Ningbo, the characteristics of the number of prenatal check-ups, time, holidays, weather and other characteristics of the previous day were introduced to establish a prediction model for the number of daily check-ups with multiple characteristics .The prediction results were compared with the classical ARIMA model. Results The average absolute percentage errors of the GBDT , RF and ARIMA model predictions were 14.95%, 17.16%, and 18.53%, respectively. Conclusions The effectiveness and feasibility of the ensemble decision tree model in medical demand forecasting is proven, and the prediction accuracy is higher than the traditional ARIMA model.

參考文獻(xiàn)：

[1] 魯翔,許年珍,袁永根,等.大型醫(yī)院醫(yī)療流程和資源配置的仿真決策系統(tǒng)研究[J]. 中國(guó)醫(yī)院管理,2005,25(01):10-13.

Lu X，Xu NZ，Yuan YG，et al. Study on medical processes and resource disposition in township and municipal hospitals: a simulation of polcy-making[J]. Chinese Hospital Management,2005,25(01):10-13

[2] 賀國(guó)光. ITS 系統(tǒng)工程導(dǎo)論［M]. 北京: 中國(guó)鐵道出版社，2004

[3] Tandberg D, Qualls C. Time series forecasts of emergency department patient volume, length of stay, and acuity[J]. Annals of Emergency Medicine, 1994, 23(2):299-306.

[4] Asplin BR, Flottemesch TJ, Gordon BD. Developing models for patient flow and daily surge capacity research[J]. Academic Emergency Medicine Official Journal of the Society for Academic Emergency Medicine, 2014, 13(11):1109-1113.

[5] Batal H, Tench J, Mcmillan S, et al. Predicting patient visits to an urgent care clinic using calendar variables[J]. Academic Emergency Medicine, 2014, 8(1):48-53.

[6] 刁秀芳,李望晨.基于SVM模型和ARIMA模型在擬合病毒性肝炎發(fā)病率中的應(yīng)用[J].現(xiàn)代預(yù)防醫(yī)學(xué),2017,44(09):1545-1548.

Diao XF, Li WC. SVM-and ARIMA-based infectious disease forecasting[J]. Modern Preventive Medicine, 2017, 44(9): 1545-1548.

[7] Gul M, Guneri A F. Forecasting patient length of stay in an emergency department by artificial neural networks[J]. Journal of Aeronautics & Space Technologies, 2015, 8(2):43-48.

[8] Yousefi M, Ferreira RP M, Yousefi M. A modeling approach for daily patient visits forecasting in an emergency department[C]// International Conference on Engineering Optimization - Iguassu Falls. Brazil：EngOpt, 2016：19-23.

[9] Qiao Z , Sun N , Li X , et al. Using machine learning approaches for emergency room visit prediction based on electronic health record data[J]. Studies in Health Technology and Informatics, 2018, 247:111-115.

[10] 鄭凱文,楊超.基于迭代決策樹(GBDT)短期負(fù)荷預(yù)測(cè)研究[J].貴州電力技術(shù),2017,20(2):82-84+90.

Zheng KW, Yang C. Research of short-term load forecasting based on gradient boosting decision tree (GBDT)[J]. Guizhou Electric Power Technology, 2017, 20(2):82-84,90.

[11] 丁聰,倪少權(quán),呂紅霞.基于梯度提升的城市軌道交通客流量預(yù)測(cè)分析[J].城市軌道交通研究,2018,21(9):60-63.

Ding C, Ni SQ, Lyu HX. Forecast and analysis of urban ｒail transit passenger flow based on gradient boosting[J].Urban Mass Transit,2018,21(9):60-63.

[12] 蔣怡玥,董蜀黔,周淑敏.基于集成算法的路段短時(shí)行駛時(shí)間預(yù)測(cè)[J].山東科學(xué),2018,31(4):118-125.

Jiang YY, Dong SQ, Zhou SM. Short term prediction of road travel time based on an ensemble algorithm[J]. Shandong Science,2018,31(4):118-125.

[13] Breiman L. Random forest[J]. Machine Learning. 2001, 45(1): 5-32.

[14] Friedman JH. Greedy function approximation: a gradient boosting machine[J].The Annals of Statistics，2001，29( 5) : 1189-1232.

[15] 李航.統(tǒng)計(jì)學(xué)習(xí)方法[M]北京:清華大學(xué)出版社,2012.

[16] Freund Y. Boosting a weak learning algorithm by majority[J]. Information and Computation. 1995, 121(2): 256-285.

服務(wù)與反饋：

【文章下載】【加入收藏】

提示：您還未登錄，請(qǐng)登錄！點(diǎn)此登錄

51黑料吃瓜在线观看,51黑料官网|51黑料捷克街头搭讪_51黑料入口最新视频