一种面向商业智能兴趣度的顾客目录分割算法

从商业智能角度出发,研究面向顾客的目录分割问题,发挥商品及顾客聚簇在交易数量中所起的作用。为平衡评分函数各部分的地位,引入了顾客对商品的兴趣度加权值,并对满足商品兴趣度的顾客数量进行评估。为优化这一评估,提出了基于顾客兴趣度的加权商品目录分割IW-BPF算法,并给出了详细的SQL算法描述和实验分析结果,验证了本算法能够获得更好的商业目标。     

客户导向目录分割问题的改进算法

客户导向目录分割问题假设顾客至少对目录中一定数量的商品感兴趣,计算目录覆盖的顾客数量,据此评估目录分割结果. 现有的分割算法为了保证目录尽可能多的覆盖顾客,而忽略了目录分割结果的效用. 针对该问题,本文构建一种新的数据存储结构CFP-Tree用于存储顾客交易数据,并提出一种新的算法Effective-Cover解决目录分割问题. 该算法使用树深度遍历法选择目录产品. 实验结果表明,该算法能够获得更好的目录分割结果.     

面向寿险客户的目录分割问题研究

基于面向顾客的目录分割问题的数学模型，通过对客户的风险评估，引入风险系数和风险阈值，将目录分割问题转化为客户与商品之间相互作用的结果。针对商品的优先权值提出一个新的计算规则，给出面向寿险客户的目录分割问题模型，适用于寿险企业的市场营销。该模型已成功应用于某保险软件公司的客户关系管理系统。     

基于偏好的客户模糊聚类方法

在电子商务环境下,如何按照顾客的购买兴趣进行聚类分析并为其提供个性化服务,是电子商务应用中研究的热点课题之一时.顾客的浏览行为及兴趣进行了研究,提出了利用偏好度的方法来度量顾客的兴趣度,在此基础上给出了基于偏好的客户群聚类算法.在该算法中,依据Web日志数据计算顾客偏好度,建立偏好度矩阵,再利用模糊聚类方法对顾客进行聚类.并用实例说明了具体的聚类过程.     

基于多尺度分析和均值漂移的Kway-Ncut算法

提出一种基于多尺度分析和均值漂移的谱聚类算法.该算法以Kway-Ncut算法为基础,通过缩小待分割图片的分辨率来实现快速和对大分辨率图片的分割.首先,利用均值漂移算法对图片进行预分割,随后缩减图像和预分割结果的分辨率.再利用预分割提供的先验信息和像素的空间一致性构建相似度模型,计算缩小后的图片像素相似度,使用Kway-Ncut进行分割.最后,将分割结果扩展为原始分辨率,用原始分辨率的预分类信息对图像边界及细节部分加以恢复,获得最终的分割结果.通过使用多幅彩色图像进行分割实验,结果表明文中算法在准确性和高效性方面都有良好表现.     

平滑度欧式聚类算法分割点云数据

提出一种平滑度欧式聚类点云分割算法,用于实现对Kinect点云的快速、准确分割.首先介绍了Kinect点云的采集和滤波方法,然后在传统欧式聚类算法基础上提出了一种平滑度欧式聚类分割算法,通过加入平滑阈值的约束来防止过度分割或分割不足的问题,并保持了较快的分割速度.通过对工业机器人获取的阀门点云数据进行实验,证明了算法的有效性.     

一种顺序无关的改进分水岭图像分割算法

为了减少顺序无关分水岭算法中的脊线标记RIDGE的数量,引入像素的湖最小值作为附加地形特征来消除不确定性。同时为了解决分水岭算法的过分割问题,引入落差来控制分割区域的形成过程。实验结果表明,改进算法在增加有限计算复杂度的情况下,将RIDGE标记数量减少了约80%,改善了原顺序无关算法中不确定像素过多的问题,将分割结果区域的数量减少了5%~20%,并且算法保持了顺序无关的特性。     

结合SSFCM与随机游走的半监督图像分割算法

针对基于颜色特征空间的半监督聚类分割算法适合分割结果包含多个颜色特征相似目标的应用场合,但对高噪声图像却无法获得理想的分割结果,而基于随机游走理论的半监督图像分割算法需要用户对目标逐一进行标记的问题,提出一种半监督图像分割算法.首先根据用户标记采用半监督模糊C均值聚类(SSFCM)算法对图像颜色特征进行建模;然后引入一个确信度函数,并根据SSFCM算法得到的隶属度数据计算确信度函数值,再将像素分为2类,分别作为随机游走图像分割算法的已标记点和未标记点;最后采用随机游走算法完成最终的分割.实验结果表明,该算法对图像中的噪声具有良好的抑制作用,且无需用户对目标逐一进行标记.     

动态相对模糊区域生长算法

为了提高区域生长的分割精度和模糊连通度算法的运算速度,减少算法所需的人工干预和种子点选取对分割结果的影响,提出了一种融合区域生长和改进模糊连通度,并结合置信区间和区域竞争方法,用于医学图像分割的动态相对模糊区域生长算法.首先算法使用置信区间区域生长快速地得出初步的多个对象分割结果,然后利用置信区间的参数,对分割区域间重叠部分使用动态相对模糊连通度算法进行再分割.通过在大量医学图像上的实验,实现了复杂背景下的图像分割.实验结果表明,该算法所需交互少,并能提高分割精度和速度.     

基于改进深度神经网络的遥感影像语义分割

研究了基于深度学习的遥感图像语义分割问题,将建筑物作为遥感图像中的待分割目标,采用语义分割算法将建筑物提取出来.提出了一种改进的U-net网络,根据分割实际需求,保持网络对目标提取特征能力的前提下,将原U-net网络的卷积核数量适当减少,降低了网络参数数量和计算复杂度;增加了Batch Normalization层抑制过拟合问题;在上采样部分增加特征图的局部信息以优化网络对于细节的分割效果.使用公开的数据集INRIA Aerial Image Dataset来评估改进的U-net网络的实际效果,和原U-net相比,单张图片训练速度提升了8％,分割精度也明显提升,训练中的过拟合情况得到改善.证明了本文改进的U-net网络具有对遥感图像的语义分割任务的有效性和可行性.     

Mining customer knowledge for electronic catalog marketing

Direct marketing is the use of the telephone and non-personal media to communicate product and organizational information to customers, who then can purchase products via mail, telephone, or the Internet. In contrast, catalog marketing is a type of marketing in which an organization provides a catalog from which customers make selections and place orders by mail or telephone. However, most catalogs for retailing firms are presented to customers in the format of paper catalogs without strategic segmentation design and implementation. In this regard, electronic catalog design and marketing could be a method to integrate the Internet and catalog marketing using market segmentation in order to enhance the effectiveness of direct marketing and sales management in retailing. This paper uses data mining based on association rules from relational database design and implementation for mining customer knowledge. As result, marketing knowledge patterns and rules are extracted for the electronic catalog marketing and sales management of a retailing mall in Taiwan.     

基于加权BPF顾客商品目录区隔挖掘算法

商品目录区隔问题是商业智能领域数据挖掘研究的一个重要问题。论文阐述了面向顾客商品目录区隔问题的最新研究成果,并提出了解决k-MECWT的加权Best-Product-Fit算法,给出了详细的SQL算法描述和应用实例。同时阐述了商品目录区隔问题的未来研究方向。     

Improving Personalization Solutions through Optimal Segmentation of Customer Bases

On the Web, where the search costs are low and the competition is just a mouse click away, it is crucial to segment the customers intelligently in order to offer more targeted and personalized products and services to them. Traditionally, customer segmentation is achieved using statistics-based methods that compute a set of statistics from the customer data and group customers into segments by applying distance-based clustering algorithms in the space of these statistics. In this paper, we present a direct grouping-based approach to computing customer segments that groups customers not based on computed statistics, but in terms of optimally combining transactional data of several customers to build a data mining model of customer behavior for each group. Then, building customer segments becomes a combinatorial optimization problem of finding the best partitioning of the customer base into disjoint groups. This paper shows that finding an optimal customer partition is NP-hard, proposes several suboptimal direct grouping segmentation methods, and empirically compares them among themselves, traditional statistics-based hierarchical and affinity propagation-based segmentation, and one-to-one methods across multiple experimental conditions. It is shown that the best direct grouping method significantly dominates the statistics-based and one-to-one approaches across most of the experimental conditions, while still being computationally tractable. It is also shown that the distribution of the sizes of customer segments generated by the best direct grouping method follows a power law distribution and that microsegmentation provides the best approach to personalization.     

Genetic algorithms for door-assigning and sequencing of trucks at distribution centers for the improvement of operational performance

In a supply chain, cross docking is one of the most innovative systems for improving the operational performance at distribution centers. By utilizing this cross docking system, products are delivered to the distribution center via inbound trucks and immediately sorted out. Then, products are shipped to customers via outbound trucks and thus, no inventory remains at the distribution center. In this paper, we consider the scheduling problem of inbound and outbound trucks at distribution centers. The aim is to maximize the number of products that are able to ship within a given working horizon at these centers. In this paper, a mathematical model for an optimal solution is derived and intelligent genetic algorithms are proposed. The performances of the genetic algorithms are evaluated using several randomly generated examples.     

Segmenting customers by transaction data with concept hierarchy

The segmentation of customers is crucial for an organization wishing to develop appropriate promotion strategies for different clusters. Clustering customers provides an in-depth understanding of their behavior. However, previous studies have paid little attention to the similarity of different items in transaction. Lack of categories and concept levels of items, results from item-based segmentation methods are not as good as expected. Through employing a concept hierarchy of items, this study proposes a segmentation methodology to identify similarities between customers. First, the dissimilarity between transaction sequences is defined. Second, we adopt hierarchical clustering method to segment customers by their transaction data with concept hierarchy of consumed items. After segmentation, three cluster validation indices are used for optimizing the number of clusters of customers. Through the compassion of normalized index, the segmentation method proposed by this study rendered better results than other traditional methods.     

Finding k most favorite products based on reverse top-t queries

A reverse top-t query for a product returns a set of customers, named potential customers, who regard the product as one of their top-t favorites. Given a set of customers with different preferences on the features of the products, we want to select at most \(k\) products from a pool of candidate products such that their total number of potential customers is maximized. Two versions of the problem are defined according to whether the competitive existing products are given. For solving this NP-hard problem, we first propose an incremental greedy approach to find an approximate solution of the problem with quality guaranteed. For further speeding up this basic greedy approach, we exploit several properties of the top- \(t\) queries and skyline queries to reduce the solution space of the problem. In addition, an upper bound of the potential customers is estimated to reduce the cost of computing the reverse top- \(t\) queries for the candidate products. Finally, when the candidate products are formed from multiple component tables, we propose a strategy to reduce the number of the accessed tuples in the component tables such that only the tuples that are possibly components of the top- \(t\) favorites of the customers need to be accessed. By applying these pruning strategies, we propose another faster greedy approach. The experiment results demonstrate that the proposed pruning strategies work very well and make the faster greedy algorithms for both versions of the problem achieve excellent performance on both efficiency and memory utilization.     

Mining changing customer segments in dynamic markets

Identifying customer segments and tracking their change over time is an important application for enterprises who need to understand what their customers expect from them – now and in the future. This in particular is important for businesses which operate in dynamic markets with customers who, driven by new innovations and competing products, have highly changing demands and attitudes. Customer segmentation is typically done by applying some form of cluster analysis to obtain a set of segments to which future customers are assigned to. In this paper, we present a system for customer segmentation which accounts for the dynamics of today’s markets. It employs an approach based on the discovery of frequent itemsets and the analysis of their change over time which, finally, results in a change-based notion of segment interestingness. Our approach allows us to detect arbitrary segments and analyse their temporal development. Thereby, our approach is assumption-free and pro-active and can be run continuously. Newly discovered segments or relevant changes will be reported automatically based on the application of several interestingness measures.     

Designing customer-oriented catalogs in e-CRM using an effective self-adaptive genetic algorithm

Analysis of customer interactions for electronic customer relationship management (e-CRM) can be performed by way of using data mining (DM), optimization methods, or combined approaches. The microeconomic framework for data mining addresses maximizing the overall utility of an enterprise where transaction of a customer is a function of the data available on that customer. In this paper, we investigate an alternative problem formulation for the catalog segmentation problem. Moreover, a self-adaptive genetic algorithm has been developed to solve the problem. It includes clever features to avoid getting trapped in a local optimum. The results of an extensive computational study using real and synthetic data sets show the performance of the algorithm. In comparison with classical catalog segmentation algorithms, the proposed approach achieves better performance in Fitness and CPU-time.     

A bilevel fixed charge location model for facilities under imminent attack

We investigate a bilevel fixed charge facility location problem for a system planner (the defender) who has to provide public service to customers. The defender cannot dictate customer-facility assignments since the customers pick their facility of choice according to its proximity. Thus, each facility must have sufficient capacity installed to accommodate all customers for whom it is the closest one. Facilities can be opened either in the protected or unprotected mode. Protection immunizes against an attacker who is capable of destroying at most r unprotected facilities in the worst-case scenario. Partial protection or interdiction is not possible. The defender selects facility sites from m candidate locations which have different costs. The attacker is assumed to know the unprotected facilities with certainty. He makes his interdiction plan so as to maximize the total post-attack cost incurred by the defender. If a facility has been interdicted, its customers are reallocated to the closest available facilities making capacity expansion necessary. The problem is formulated as a static Stackelberg game between the defender (leader) and the attacker (follower). Two solution methods are proposed. The first is a tabu search heuristic where a hash function calculates and records the hash values of all visited solutions for the purpose of avoiding cycling. The second is a sequential method in which the location and protection decisions are separated. Both methods are tested on 60 randomly generated instances in which m ranges from 10 to 30, and r varies between 1 and 3. The solutions are further validated by means of an exhaustive search algorithm. Test results show that the defender's facility opening plan is sensitive to the protection and distance costs.