【摘要】：近年來(lái)人們不斷的研究人口發(fā)展的規(guī)律，希望能從復(fù)雜多變的人口中找到一個(gè)規(guī)律來(lái)預(yù)測(cè)人口未來(lái)的發(fā)展，從而制定合理的政策。但人口的增長(zhǎng)易受出生率，死亡率等客觀因素和人口政策等主觀因素的影響，所以一般傳統(tǒng)的方法對(duì)于人口的預(yù)測(cè)精度往往達(dá)不到所期望的數(shù)值。而人工神經(jīng)網(wǎng)絡(luò)是一門非線性科學(xué)，具有很強(qiáng)的容錯(cuò)性，非線性的映射能力和自適應(yīng)性，可以使用非線性映射表示人口數(shù)量這一非線性系統(tǒng)用以提高模型精度，使它在神經(jīng)系統(tǒng)方面，模式識(shí)別，組合優(yōu)化，預(yù)測(cè)等領(lǐng)域有了成功地應(yīng)用。 本文采用三類人工神經(jīng)網(wǎng)絡(luò)：反向傳播網(wǎng)絡(luò)、RBF神經(jīng)網(wǎng)絡(luò)、時(shí)間序列預(yù)測(cè)法，研究人口預(yù)測(cè)，旨在人口預(yù)測(cè)的特征，綜合考慮人口預(yù)測(cè)的各個(gè)指標(biāo)，從而合理預(yù)測(cè)人口的增長(zhǎng)數(shù)量，，為我國(guó)的可續(xù)發(fā)展提供便利。在BP網(wǎng)絡(luò)中，為了避免網(wǎng)絡(luò)陷入局部最小點(diǎn)和提高網(wǎng)絡(luò)的收斂速度，采用動(dòng)量法與學(xué)習(xí)速率自適應(yīng)調(diào)整相結(jié)合的算法，對(duì)于全國(guó)人口總?cè)丝诘念A(yù)測(cè)，采用三層BP神經(jīng)網(wǎng)絡(luò)，其中輸入層神經(jīng)元的個(gè)數(shù)為8，輸出層神經(jīng)元個(gè)數(shù)為1.而在RBF神經(jīng)網(wǎng)絡(luò)中采用的參數(shù)是基函數(shù)的中心和方差以及權(quán)值,在時(shí)間序列模型中，采用曲線擬合和參數(shù)估計(jì)方法(非線性最小二乘法)對(duì)網(wǎng)絡(luò)進(jìn)行訓(xùn)練。對(duì)于影響全國(guó)人口總量的各個(gè)指標(biāo)也進(jìn)行了預(yù)測(cè)，建立了BP網(wǎng)絡(luò)，RBF網(wǎng)絡(luò)和AR模型的預(yù)測(cè)。 通過(guò)選取1990-2008年的人口指標(biāo)進(jìn)行預(yù)測(cè)，預(yù)測(cè)結(jié)果表明，總?cè)丝跀?shù)量預(yù)測(cè)值與實(shí)際值基本吻合，BP預(yù)測(cè)值與人口總量誤差0.0046,0.0011,0.0009,0.0035,0.0000。RBF預(yù)測(cè)值與人口總量誤差為：0.0012,0.00023,0.0062,0.0141，0.0056。AR模型預(yù)測(cè)值與人口總量誤差：0.0031,0.0045，0.0079,0.0002，0.0005。對(duì)于其他指標(biāo)的預(yù)測(cè)，三種網(wǎng)絡(luò)的預(yù)測(cè)值與實(shí)際值也非常接近，從而說(shuō)明神經(jīng)網(wǎng)絡(luò)用于人口的預(yù)測(cè)是可行和有效的，有著良好的前景。
[Abstract]:In recent years, people have been constantly studying the law of population development, hoping to find a law from the complex and changeable population to predict the future development of the population, so as to formulate reasonable policies. However, the population growth is easily affected by objective factors such as birth rate, death rate and population policy. The artificial neural network is a nonlinear science, which has strong fault tolerance, nonlinear mapping ability and adaptability. It can be used to improve the accuracy of the model by using nonlinear mapping to represent the population, which is a nonlinear system. It has been successfully applied in the fields of nervous system, pattern recognition, combinatorial optimization and prediction. In this paper, three kinds of artificial neural networks are used: back propagation network, RBF neural network, time series forecasting method. Thus the reasonable forecast population growth quantity, provides the convenience for our country's sustainable development. In BP network, in order to avoid the network falling into the local minimum point and improve the convergence rate of the network, the momentum method is combined with the adaptive learning rate adjustment algorithm, and the three-layer BP neural network is used to predict the total population of the whole country. The number of neurons in the input layer is 8 and that in the output layer is 1. The parameters used in RBF neural network are the center, variance and weight of the basis function. In the time series model, the methods of curve fitting and parameter estimation (nonlinear least square method) are used to train the network. The BP network, RBF network and AR model are established. By selecting the population index from 1990 to 2008 to forecast, the forecast result shows that the predicted value of the total population quantity basically coincides with the actual value. The forecast value of BP and the total population error are 0.0046C 0.00110.0009U 0.0035N 0.0000.RBF forecast value and total population error: 0.00120.00023N 0.0062n00141N 0.0056.AR model forecast value and population gross error: 0.0031n0045c0.00790.00020.0005. For the prediction of other indexes, the predicted values of the three networks are very close to the actual values, which shows that the neural network is feasible and effective in population prediction, and has a good prospect.
【學(xué)位授予單位】：中北大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2012
【分類號(hào)】：O212;C921

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 陳捷;陳為真;;灰色神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)修正模型研究[J];武漢工業(yè)學(xué)院學(xué)報(bào);2011年03期

2 戴添發(fā);趙鴻;;隨機(jī)變異—優(yōu)化選擇神經(jīng)網(wǎng)絡(luò)在鐵貨運(yùn)量預(yù)測(cè)方面的應(yīng)用[J];新疆師范大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年02期

3 欒清華;劉德東;李永濤;石萍;王曉琳;;閩江十里庵洪水預(yù)報(bào)人工神經(jīng)網(wǎng)絡(luò)模型試驗(yàn)[J];灌溉排水學(xué)報(bào);2007年S1期

4 李倩;胡邦輝;王學(xué)忠;顧錦榮;;基于BP人工神經(jīng)網(wǎng)絡(luò)的區(qū)域溫度多模式集成預(yù)報(bào)試驗(yàn)[J];干旱氣象;2011年02期

5 丁亞;;模擬電路故障診斷技術(shù)研究[J];中國(guó)科技信息;2011年18期

6 馮羽;馬鳳山;魏愛華;趙海軍;郭捷;;灰色系統(tǒng)與神經(jīng)網(wǎng)絡(luò)組合模型在地下水水位預(yù)測(cè)中的應(yīng)用[J];中國(guó)地質(zhì)災(zāi)害與防治學(xué)報(bào);2011年03期

7 姜偉;馬令勇;劉功良;;基于MATLAB神經(jīng)網(wǎng)絡(luò)方法的多層磚房震害預(yù)測(cè)[J];西北地震學(xué)報(bào);2011年02期

8 楊東輝;馬良;黃衛(wèi)東;;基于人工神經(jīng)網(wǎng)絡(luò)的激光立體成形件成形表面質(zhì)量預(yù)測(cè)[J];中國(guó)激光;2011年08期

9 姚寧s

本文編號(hào)：2382578