论文研究Multiobjective Optimization Algorithms for PointofInterest Route Planning in LBSN.pdf

LBSN中兴趣点路径规划的多目标优化算法，许莹，郭昱杉，基于位置的社交网络(LSBN)为用户提供平台，以在便兴趣点(POI)上共享位置，照片或评论。如何挖掘这些共享信息以向用户推荐兴趣点或旅�
山国科技论文在线 http:/www.paper.edu.cn Then gcnctic algorithm is uscd to solve the problcm In this paper, we firstly model the POl route recommendation problem as a multiobjective oplimization problem with three objectives, including the user preference, the location popularity, and the travel distance. then the user checkin dataset from lbsn is used to mine user 95 personalized information and obtain thc uscrs prcfcrrcd POls at ccrtain pcriod of timc. By using collaborative filtering, the Topk POls can be obtained as the candidate recommendation POls from each POI set. Next, four multiobjective route recommendation algorithms based on evolutionaryalgorithms, including MOEA/DPR MOEA/d based POl Route), NSGAIIPR (NSGAII based POIRoute ), SPEAIIPR(SPEAIl based POiRoute), PESAlIPR(PESAII 100 based POIRoute) have been designed to obtain a set of nondominated Pareto solution which containing multiple poi travel routes. The performance of these four multiobjective route planning algorithms has been investigated and compared. Finally, experimental results demonstrate that our proposed algorithms have better performance compared with some existing algorithms with rcspcct to thc accuracy and diversity 105 The lbsn data from JiePang are used in this work. The dataset d=data,, data2, data3, data4, datas) represents user checkin records. datai =< u, t,l> is the ith checkin dala. where u is the user ID, l is the checkin Lime, l is the location, i.e a POI. For each 1, a vector <l(x, y), 1namc, 1  type, 1c>is defined, where l(x, y) is 110 coordinates of I in which x represents the longitude and y represents the latitude, Iname is the location name, Itype is the location type Here location type =(V,F, e).IcE typei is the subcategory of type For example, Venue can be schools, office buildings, scenic buildings, etc. Spor is thc sct of POIS, SPoI=(1,2,,i.,I rcprcscnts thc ith POI in Spor Route=(1,I2,.,i,. is the planned sequence of POls, i. e. a POi route, where each Iiis 115 selected from a POi set for a given time slot. In order to define the multiobjective POi route planning problem, some key notations are applied and described in Table 1 In this work we define the multiObicctivc POI Route Planning( mopoirp)problcm which concerns three objectives, including maximizing the popularity of POl, maximizing the location popularity, and minimizing the total travel distance as follows 120 1) Maximize user preference By calculating the visiting frequency of each location category at different time slots, the highest one at different time slots is regarded as the user preferred category at different time slots According to the previous results, all the Pols in St belong to the preferred category at time slot tk. In addition, St, is obtained by deleting locations with the distance d greater than dmar. Topk 125 POIs from Sti are selected as the recommended POI set St, we calculate the user preference F(u, typei, C;, tk) for the subcategory c in typei at a given time slot tk as follows ∑!cec;cont(u!;tk,) F(u, typei, Ci, tr) ount(u, Lk, l) where j=1, 2 ,.typeil, Itypeil is the number of categories for typei. The preference Ftk., of u to typei at time slot tk can be defined in Formula(2) 130 K F(u, typei, Cl,tk),, F(u, typei, Cn,tr)>(2) Where n=typel, F(u, type;, Cn,tk) for c; E type is calculated based on Formula(1) The preference of u for location I at time slot tk, denoted by like(u, tk, l can be calculated as 山国酬技论文在线 http:/www.paper.edu.cn follows lke(x,t,)=、a(D.(3) 135 Because of the sparsity of user's checkin data, the user based collaborative filtering is used to estimate the user preference. Sim(u, 1, cj, tk) is defined as the similarity between user u and user v for location category Ci at time slot tk using the cosine similarily as follows Sim(u, V,Ci, tx) (4) (u, tk, L) is delined as user's preference of location I at time slot tk, where U(u, cj) is the set of similar users of u for location category c; R(u, tk,l) EucU(uc) Sim(u, v, c i. tk)like(v. tk D) We can obtain all preferred locations in St, using the above formulas. S'ti is formed by selecting Topk POls from S Tab. 1 Key notations Symbol Dcscription D checkin Data sets S the set pol location represented b <longitude, latitude L type location typc C the maximum distance that user can visit ar the time limit d(l1,l2) the distance between, and l2 stay(l) stay time at location l T(l1,l2) he travel time froml, and C the set of location categories checkin time the kth time sl Route the Pol route, i.e. a sequence of ordered POls type =(V, F,E location types, V for Venue, F for Food, E for Entertainment The poi set for by deleting locations with th distance d grcatcr than dmax S Topk POls from St, are selected as the recommended pol 145 2) Maximizelocationpopular Location popularily is limedependent. So, we calculate the popularity of all POis in St,. lor a certain month. Therefore, a time factor m is introduced to calculate the popularity H: H(,m)= VT(L,m) VT(L,m) VT (L, m) represents the visited times for location I in the mth month 150 Based on the above definition, mOPOiRP problem considers three objectives, including maximizing the total preferences (i.e. the sum of all Poi preferences in the recommended route) maximizing the total popularity (i.e. the sum of all Poi popularity)and minimizing the total travel distance (i.e. the sum of all distances in the route), can be defined as follow 4 山国科技论文在线 http:/www.paper.edu.cn f(0 Maximize the total user preference fi(D=Max >like(Li) 155 f2(1)Maximize the total popularity f2(=Me L f3(1)Minimize the total travel distance f3(0)=Min>d(1,4+1)+d(lo, L1) S. t d=ddo, li< dmax tatt tma 160 Here t1=2i1 stay(l1), t2=2i=o T(, i+i) n= Routel Based on the wellknown MOEA/D, we propose MOEA/DPR(MOEA/D based POI Route) to solve the mopoirP problem concerned in this paper. MOEA/d is a multiobjective 1 65 evolutionary algorithm combining mathematical programming with evolutionary algorithm. It decomposes the MOP problem into a number of scalar optimization subproblems and then optimizes them simultaneously. To solve the MOPOirP problem, the populalion initialization and update process is designed as shown in Fig. 1. A route of POIs is a chromosome. Each gene in the chromosome corresponds to a location in the route. Fig. 1(a) represents a chromosome with five 170 genes that correspond to different locations. Fig. I(b)shows locations are randomly selected from different st. to construct a chromosome. Fig. I(c) shows four random initial routes have been generated. The basic operation of MOEA/DPR is similar to MOEA/D. MOEA/DPR is then designed to generate a set of nondomination solutions, i. e. POI routes, for the mOPoirP problem e( location1)(s+(S2→(Sn→ →tXSs)1) Routel fcnc(locati l2 4→M1XS→)l) Route iene( location Gene( locations) →+(S2…S→+)(sS=) Routed S 〔b) 0n1 cations 175 山国科技论文在线 http:/www.paper.edu.cn Fig 1: The procedure of generating the POi route NSGAII adopts the fastnondominance sorting algorithm, the congestion comparison operator and the elite strategy to improve the computation speed and robustness. Similar to the 80 MOEA/DPR algorithm, NSGAIlPR (NSGAII bascd POI Route) is proposcd to salve thc MOPOIRP problem. After the population initialization (see Fig. 1), the population is sorted according to the domination relation. For each generation, the offspring population is obtained through the roulette selection, crossover(Fig. 2(a)) and mutation (Fig. 2(b)). Based on the nondomination sorting, optimal solutions are added into the nondominant solution set. This 1 85 procedure repeats until the termination condition is met. The set of POI routes are the nondominated solution set obtained at the end of the algorithm SPEAII is the improved strength Pareto evolutionary algorithm. which incorporates a finegrained fitness assignment strategy, a density estimation technique and an enhanced archive 190 truncation method. SPEAII has good performance in terms of the convergence and diversity Samc as the previous mcthod, SPEAlIPR(SPEAlI bascd POI Route)is proposcd to solvc the MOPOiRP problem. The genetic operation of the population is shown in Fig. 2 S→2)S)Sn→b8S=)(S+)(5:(Sx)(S×)(Sn 5,…→1)(8→ 间a) The crossover 195 )(SxA14)(S4小)(S Mutation (Si> 3 iS)isIs yip (b) The mutation Fig 2: The Evolutionary Operators A PESAlI based algorithm PESAllPR(PESAlI based POI Route) is proposed to solve the MOPOIRP problem, where PESAII is the improved ParetoEnvelope based Selection Algorithm Instead ofassigning the fitness lo an individual, the fitness in PESAlI is assigned to hyperboxes in the objective space. The objective space is divided into many boxes. Based on the population 205 initialization shown in F PESAIIPR also uscs Ga to update the population. In ordcr to reduce the computational cost, PESAlI uses the regionbased selection to select points(solutions) in hyperbox within the objective space We used the lbsn dataset of user checkin records obtained from JiePang, where each entry 210 contains user ID, checkin time, checkin geographic coordinates, location name, location category and location type. We randomly select a set of 4248 anonymous users with at least 100 checkin 山国酬技论文在线 http:/www.paper.edu.cn records. Thc datasct includes a total of 1263 32 records Wc sct 80%o chcckin records as thc training data. and the rest is used as the test data 215 Hitrate and Diversily are used to measure experimental resulls, which can be describedasfollow: I The hitrat HitrateEuculR(u)nT(u(7) Here, R(u is the number of different POIs in the recommended travel packages of user u 220 R(un T(uI is the number of recommended POIs, which appear in the verification set t(u) of Iser u 2) The Diversity A recommendation should provide diverse recommendation results to users. The diversity of recommendation result for user u is defined as 225 ∑1cR( uj, ii Similarity(ll Diversity= 1 R(tD)(R(u)1) (8) Here, Similarity (li, li)means the similarity between location li and lj. If location li and lj belong to the same category, Similarity (i, li) is set as 1; otherwise, Similarity (li, lj) is set as 0 3)Fmeasure 230 Fmeasure combines precision and recall together with a harmonic mean. Here we use commonly used FB with B=1, i.e. F1. If the value of F1 is high, it means that the test method is eftective FB=(1+B2) PrecisionxRecall B2Precision+Recall Here, Precision =R(u)nT(uI/R(ul, Recall=R(u)nT(u/R(u), where T(u)are the 235 different POIs a user visited in the verification set and rlu) are the different POis in the recommended routes In order to determine the appropriate iterations for these four multiobjective algorithms, the Hitrate of these four algorithms is compared by setting different iterations from 50, 100 to 200 in Fig 3. According to the test result, these four algorithms can obtain better or similar results when 240 the number of iteration is set as 50. We calculate diversity defined in Formula(7)of these four multi objcctivc algorithms, which show that all valucs arc distributcd in the rangc of (0.8, 1)as shown in Fig 4. It indicates that using the proposed algorithms can generate the diverse pol routes i.e. these pois belong to different categories. Thus, we set the number of iterations as 50 in all proposed algorithms In this group of experiments, the hitrate of four multiobjective POI route planning algorithms is comparcd in Fig 3. Hcrc, the horizontal axis indicates thc test group numbcr and thevertical axis represents the corresponding hit rate values. We can see that, for the MOPOiRP problem, our proposed NSGAlIPR obtained the highest Hitrate compared with other algorithms 250 From Fig. 4, we can see that the four multiobjective optimization algorithms have similar diversity besides, the F1measure value of nsgaIlPr is also better than other three algorithms as shown in Fig. 5 7 山国酬技论文在线 http:/www.paper.edu.cn 0.9 0.8 7 0.0000 6543 0.2 0.1 BNSGAIIPR OSPEAIIPR EPESAIIIIPR GMOEA/DPR Fig 3: The comparison of Hitrate △ 0000 9876 △ 6 7 NSGAILPR eCO. SPEAILPR 255 一x一 PESAIIPR 尘=MOEA/DPR Fig 4: Comparison of Diversity 000 876 0.5 0.4 3210 NSGAlPR國SPEA PR PESAllPR口 MOEA/DPR Fig 5: The comparison ofFimeasure Experimental results in 9I showed that the performance of tRP is better than the random 260 method and the hottest method. Since NSGAIIPR and SPEAlIPR have better performance among our proposed four multiobjective POi route planning algorithms, so we compare them with the statcofart algorithm trP in the litcraturc. fig. 6 shows nsgalIPR improves hitrate by 10% compared with TRP. SPEAllPR also obtained better Hitrate on two out of the three test data sets. Fig. 7 shows the comparison of Frmeasure. NSGAllPR obtained better FImeasure 265 results than other algorithms. Experimental results demonstrate that our proposed algorithms are competitive approaches for solving the moPoirP problem in LBSN 山国酬技论文在线 http:/www.paper.edu.cn 0 86420 6 7 SNSGAIPR O SPEAIIPR TRP Fig 6: Comparison of Hitrate with TRP 000 0000 87654321 LONSGALILPR 3 ESPEAJILPR TRP 7 270 Fig. 7: Comparison of Flmeasure with TRP In this papcr, thc POi route planning problcm is modeled as a multiobjcctivc optimization problem by considering three objectives, including maximizing the user preference, maximizing the location popularity and minimizing the travel distance. Four new multiobjective optimization 275 algorithms including NSGAIIPR, SPEAIIPR, PESAIlPR, and MoEa/DPR have been proposed for thc first timc to solvc the MultiObjcctivc Optimization Poi route planning problcm Experiments on some real lbsn data set demonstrate SPEAllPR obtained the overall best nondominated solution set among the proposed four multiobjective Pol route planning algorithms. When considering the hit rate of recommendation, NSGA1IPR performed the best compared with other three multiobjective Poi route planning algorithms proposed in this paper and the trp algorithm in the litcraturc. In the future, we plan to further improve the multiobjective optimization algorithms by applying some machine learning approaches to solve the poi route planning problem 285 [1]Bao J. Zheng Y, and Mokbel M F Locationbased and preferenceaware recommendation using sparse geosocial networking data [A]. In Proceedings of International Conference on Advances in Geographic Information Systems[C]. 2012. 199208 [2] Kojima N, Takagi T. Location Recommendation Incorporating Temporal and Spatial Effects[A]. In International Conference on Web Intelligence and Intelligent Agent Technology[C]. 2016. 280283 290 [31 Yu Z, Feng Y, Xu H, and Zhou X. Recommending travel packages based on mobile crowdsourced data[ J1 IEEE Communications Magazine, 2014, 52(8): 5662 [4 Liu B, Xiong H, Papadimitriou S, et al. A General Geographical Probabilistic Factor Model for Point of Interest Recommendation[J]. IEEE Transactions on Knowledge Data Engineer ering,2015,27(5):11671179 [5] Cheng C. Yang H, King I. et al. Fused matrix factorization with geographical and social influence in 295 locationbased social networks[A]. AAAI Conference on Artificial Intelligence[C]. 2012 [6 Zhang Q, and liH MOEA/ D: A Multiobjective Evolutionary Algorithm Based on Decomposition[J]. IEEE Transactions on Evolutionary Computation, 2007, 11(6): 712731 .7 Deb K, and Jain IL. An Evolutionary ManyObjective Optimization Algorithm Using ReferencePointBased Nondominated Sorting Approach, Part 1: Solving Problems With Box Constraints[J]. IEEE E Transactions on 300 Evolutionary Computation, 2014, 18(4): 577601 山国酬技论文在线 http:/www.paper.edu.cn [8]Xu Y, Ding O, Qu R, and Li K Hybrid multiobjective evolutionary algorithms based on decomposition for wireless sensor network coverage optimization[]. Applied Soft Computing, 2018, S1568494618301868 Pareto approach []. IEEE Transactions on Evolutionary Computation, 1999, 3(4): 2572Y an [9]Zitzler E and Thiele L. Multiobjective evolutionary algorithms: a comparative case study and the strength 305 [10] Wang H G, and Liang M A Multiobjcctivc particlc swarm optimization[J]. Computcr Enginccring Applications, 2008, 45(4): 8285 [11Varadharajan T K, and Rajendran C A. Multiobjective simulatedannealing algorithm for scheduling in flowshops to minimize the makespan and total flowtime of jobs [ J. European Journal of operational Research, 2005,167(3):772795 310 2] Corne D w, JerramNR, Knowles J D, et al. PESAII: Regionbased Selection in Evolutionary Multiobjective A Conference on Genetic Evolutionary Computation(c]. 2001, 283290 [13]Zheng Y. Tutorial on LocationBascd Social Nctworks [A]. In proceeding of the 2lst International confcrcncc on World wide Web[]. 2012 14] Peng D, and Zhang C. Research on Intelligent Route Programming for DIY Travel[A]. In Proceeding of 315 Intermational Conference on Computer Sciences and Applications[c]. 2014. 35035 5] Bao J, Zheng Y, Wilkie D, Mokbel M. Recommendations in locationbased social networks: a survey[] Geolnformatica, 2015, 19(3): 525565 16 Cheng C, Yang H, King l, et al. A Unified PointofInterest Recommendation Framework in LocationBased Social Networks[. ACM Transactions on Intelligent Systems Technology, 2016,8(1):10 320 [17] Xu G, Fu B, and Gu Y. PointofInterest Recommendations via a Supervised Random Walk Algorithm[J] EEE Intelligent Systems, 2016, 31(1): 1523 8]Liu Y, Liu C, Liu B, et al. Unified PointofInterest Recommendation with Temporal Interval Assessmen[A] In Proceeding of ACM SIGKDD Intemational Conference on Knowledge Discovery and Data Mining[C]. 2016 10151024. 325 [19] Lin S. Computer solutions of the traveling salesman problem[J]. Bell Labs Technical Journal, 2014, 44(10) 22452269 [20]Yu Z, Xu h, Yang Z and Guo B Personalized travel package with multi pointofinterest recommendation based on crowdsourced user footprints[J]. IEEE Transactions on HumanMachine Systems, 2015, 18 [21]Kenteris M, Gavalas D, Pantziou G, Konstantopoulos C Nearoptimal personalized daily itineraries for a 330 mobile tourist guide[A]. Computers and Communications[C]. 2015. 862864 [22]Wu P, Zhu J, and Zhu L. Rcscarch on the travel route dcsign based on multiobjectivc planning and intelligent optimization algorithm J Mathematics in Practice and Theory, 2016. (15): 105114 [23]Chen C; Zhang D, (iuo B, et al. Trip Planner: Personalized Trip Planning Leveraging Ileterogeneous Crowdsourced Digital Footprints[J]. IEEE Transactions on Intelligent Transportation Systems, 2015, 16(3) 33512591273 [24]Han Y, Guan H, Duan J Tour Route Multiobjective Optimization Design Based on the Tourist Satisfaction(J. Discrete Dynamics in Nature and Society, 2014(1): 18 [25 Hasuikc T, Katagiri H, Tsubaki H, ct al. Routc planning problcm with groups of sightsccing sitcs classified by tourist's sensitivity under TimeExpanded Network[A]. IEEE International Conference on Systems, Man and 340 Cybernetics[C]. 2014. 188193 [26]Zhang Y J and Xu K L MultiObjective Team Personalized Travel Design Based on Fuzzy APACA[I Computer Engineering and Applications, 2012, 48(35): 207212. 345 中兴趣点路径规划的多目标优化 算法 410006 摘要 (LSBN) (POn 350 355关键词 LSBN POI 中图分类号:TP39 10
 874KB
论文研究LBSN中考虑用户交友偏好的好友推荐方法研究.pdf
20190920论文研究LBSN中考虑用户交友偏好的好友推荐方法研究.pdf, 基于位置的社交网络（locationbased social networks，LBSN）大为流行之余，也带来了信息过载问题.好友
 33.35MB
Book.(2001)MultiObjective_Optimization Using Evolutionary Algorithms
20181006Book.(2001)MultiObjective_Optimization Using Evolutionary Algorithms使用进化算法解决多目标问题，英语原版经典，绝版图书
 277KB
论文研究A parallel multiobjective evolutionary algorithm based on decomposition in biobjective spaces.pdf
20190816二目标空间中的一种并行分解型多目标进化算法，刘靖伟，应伟勤，分解型多目标进化算法(MOEA/D)借用了传统的数学分解方法。MOEA/D在解决多目标优化问题(MOPs)时取得了较好的实验结果，但是当解决
 503KB
Using Engineering View to Design Better Multiobjective Evolutionary Algorithms
20191230Using Engineering View to Design Better Multiobjective Evolutionary Algorithms，郑波尽，李元香，MultiObject
 2.81MB
cec 2013 代码及论文
20180527Single objective optimization algorithms are the basis of the more complex optimization algorithms s
 586KB
Introduction to Global Optimization
20130725Accurate modelling of realworld problems often requires nonconvex terms to be introduced in the mod
 975KB
k2011003.pdf
20200430MultiObjective Optimization Using Evolutionary Algorithms: An Introduction,关于进化多目标优化算法的介绍的一篇论文
 24.72MB
Handbook of Research on Soft Computing and NatureInspired Algorithms
20180619Soft computing and natureinspired computing both play a significant role in developing a better und
 27.2MB
Machine.Learning.Optimization.and.Big.Data
20161230Machine Learning, Optimization, and Big Data: Second International Workshop, MOD 2016, Volterra, Ita
 65.90MB
Artificial.Intelligence.and.Soft.Computing.17th.International.Conference.Part.I
20180530各种新算法研究，没有足够的数学和英语基础就别下载了。 Artificial Intelligence and Soft Computing: 17th International Conference
 17.28MB
Intelligent.and.Evolutionary.Systems.3319490486
20161118Intelligent and Evolutionary Systems: The 20th Asia Pacific Symposium, IES 2016, Canberra, Australia
 4.78MB
Introduction_to_Optimum_Design.pdf
20081118Preface ix Chapter 1 Introduction to Design 1 1.1 The Design Process 2 1.2 Engineering Design versus
 12.79MB
Research Advances in Cloud ComputingSpringer(2017).pdf
20180109Cloud computing is a novel computing paradigm which has changed the way enterprise or Internet compu

weixin_39841365
等级：
关注 私信 TA的资源

下载
大型冷却塔控制爆破技术及危害控制措施
大型冷却塔控制爆破技术及危害控制措施

下载
救生舱内部流场的CFD模拟
救生舱内部流场的CFD模拟

下载
裂隙带布置高抽巷实现巷道快速掘进
裂隙带布置高抽巷实现巷道快速掘进

下载
禹州华夏植物群与含煤地层研究
禹州华夏植物群与含煤地层研究

下载
磁化水改善混凝土抗渗性能试验研究
磁化水改善混凝土抗渗性能试验研究

下载
GeoDemo20200709.zip
GeoDemo20200709.zip

下载
直接甲醇燃料电池的模糊PID控制研究
直接甲醇燃料电池的模糊PID控制研究

下载
iOS 原生加载微信h5支付链接跳转微信方法封装，更新WKWebView的使用，不能用就来找我
iOS 原生加载微信h5支付链接跳转微信方法封装，更新WKWebView的使用，不能用就来找我

下载
基于一元非线性回归理论的矿井深部瓦斯含量预测
基于一元非线性回归理论的矿井深部瓦斯含量预测

下载
iOS 原生加载支付宝支付链接跳转支付宝方法封装
iOS 原生加载支付宝支付链接跳转支付宝方法封装

下载
Python实现：2019年河北省城镇就业人员年平均工资的单样本t检验.docx
Python实现：2019年河北省城镇就业人员年平均工资的单样本t检验.docx

下载
怀菊花中总黄酮的超声波提取工艺研究
怀菊花中总黄酮的超声波提取工艺研究

下载
GB 4943.12011信息技术设备 安全 第1部分：通用要求.pdf
GB 4943.12011信息技术设备 安全 第1部分：通用要求.pdf

下载
H264toPS.zip
H264toPS.zip

下载
甲级资质测绘申请所需资料（精简版）.docx
甲级资质测绘申请所需资料（精简版）.docx

下载
煤层底板灰岩含水层突水危险性评价
煤层底板灰岩含水层突水危险性评价

下载
数字图像输出分辨率对照表(ppi).png
数字图像输出分辨率对照表(ppi).png

下载
归海数据桌面全文搜索软件
归海数据桌面全文搜索软件

下载
修正的共轭梯度法在两种线搜索下的全局收敛性
修正的共轭梯度法在两种线搜索下的全局收敛性

下载
yunshixun.exe
yunshixun.exe

下载
直埋玻璃钢聚氨酯保温管及其保温层厚度设计
直埋玻璃钢聚氨酯保温管及其保温层厚度设计

下载
联想c8100打印机驱动 v1.0 官方版
联想c8100打印机驱动 v1.0 官方版

下载
学生身高和体重的假设检验及python实现.doc
学生身高和体重的假设检验及python实现.doc

下载
究极绿宝石3永久mega开发.rar
究极绿宝石3永久mega开发.rar

下载
分类器集成在入侵检测中的应用研究
分类器集成在入侵检测中的应用研究

下载
爱普生r800打印机驱动 v7.5.6.SC 官方版
爱普生r800打印机驱动 v7.5.6.SC 官方版

下载
bf5_new.exe
bf5_new.exe

下载
高压制备Ce掺杂Ca3Co4O9陶瓷的热电性能
高压制备Ce掺杂Ca3Co4O9陶瓷的热电性能

下载
等保2.0详细讲解内容以及与1.0的变化.docx
等保2.0详细讲解内容以及与1.0的变化.docx

下载
吐哈盆地沙尔湖地区煤层气钻完井适用技术探讨
吐哈盆地沙尔湖地区煤层气钻完井适用技术探讨