【摘要】：提出了一種非負(fù)矩陣分解-自回歸模型,并用該模型對居民出行流量進行預(yù)測.該模型首先利用非負(fù)矩陣分解方法挖掘城市區(qū)域內(nèi)的居民出行特征,而后在非負(fù)矩陣分解獲得的特征矩陣和系數(shù)矩陣基礎(chǔ)上對時序系數(shù)矩陣建立自回歸模型,進而對起訖矩陣進行預(yù)測.以北京市出租車數(shù)據(jù)為基礎(chǔ),與時空權(quán)重K近鄰、傳統(tǒng)K近鄰、反向神經(jīng)網(wǎng)絡(luò)、樸素貝葉斯、隨機森林和C4.5決策樹回歸模型對比,實驗結(jié)果表明,該模型的預(yù)測準(zhǔn)確率有顯著提升.
[Abstract]:A non negative matrix decomposition autoregressive model is proposed, and the model is used to predict the travel traffic. The model first uses the nonnegative matrix decomposition method to excavate the residents' travel characteristics in the urban area, and then builds the autoregressive model of the time series coefficient matrix on the basis of the characteristic matrix and the coefficient matrix obtained by the nonnegative matrix decomposition. On the basis of Beijing taxi data, it is compared with the time and space weight K nearest neighbor, the traditional K nearest neighbor, the reverse neural network, the naive Bayesian, the random forest and the C4.5 decision tree regression model. The experimental results show that the prediction accuracy of the model has been significantly improved.
【作者單位】： 西南大學(xué)計算機與信息科學(xué)學(xué)院;
【基金】：國家自然科學(xué)基金項目(61403315,61402379)
【分類號】：U491.12
，

本文編號：2121547