Kun-Nyeong Chang

Individual and group behavior-based customer profile model for personalized product recommendation

The development of efficient customer profile models is crucial for improving the recommendation quality of the recommendation system. In this paper, we propose a new customer profile model based on individual and group behavior information such as clicks, basket insertions, purchases, and interest fields. We also implement a recommendation system using the proposed model, and evaluate the recommendation performance of the proposed model in terms of several well known evaluation metrics. Experimental results show that the proposed model has a better recommendation performance than existing models.

Efficient distributed power control for cellular mobile systems

Various distributed carrier-to-interference (C/I) balancing procedures have been developed to cope with implementational drawbacks of centralized power-control schemes. However, it would be inevitable for these distributed schemes to lose some system performance compared to a centralized one since they control their power level based on only the local C/I measurements. An efficient distributed power-control (DPC) scheme is proposed to maximize the system performance. A distributed C/I balancing algorithm is suggested, which converges fast and is robust to C/I estimation errors. A simple instability detection rule is also suggested to accelerate the balancing phase of DPC schemes. Numerical results indicate that our algorithm achieves improvements in terms of the outage probability as well as the algorithm speed. Robustness to the C/I measurement errors is also explored.

An efficient borrowing channel assignment scheme for cellular mobile systems

In this paper, we propose a new efficient borrowing channel assignment (BCA) scheme, which consists of two phases. The first ordinary channel allocation phase borrows a channel from neighboring cells by an impact-based borrowing strategy. The second channel reallocation phase has a reallocation procedure for locked-channel utilization and a reallocation procedure for efficient channel reuse. Simulation results show that in both uniform and nonuniform traffic cases, our schemes significantly reduce the system blocking probability over the other existing schemes. Furthermore, one of our schemes has a much smaller number of reallocations than other compared schemes.

A real-time algorithm for timeslot assignment in multirate return channels of interactive satellite multimedia networks

Since the digital video broadcast-return channel via satellite (DVB-RCS) standard was released in 2000, developing an interactive satellite multimedia (ISM) network has become a hot issue. In order to provide high-speed multimedia services using a DVB-RCS system, it is important to efficiently and dynamically assign the timeslots to a number of terminals according to their various demands. Also, it is imperative to improve the degradation of link quality due to rain-fade attenuation in Ka-band satellite communications. As a result, multirate superframe structures should be implemented for a relatively low data rate at the cost of stable connection to terminals in rain-fade regions and a relatively high data rate to terminals in clear-sky regions. Timeslot scheduling in this environment is studied in this paper. We mathematically formulate the timeslot assignment problem as a nonlinear integer programming problem and develop an efficient real-time solution algorithm. Extensive simulation results show that our algorithm successfully finds a feasible solution with optimality gap less than 0.05% within about 5 ms at Pentium III PC. We believe that our algorithm can be utilized as a guideline in developing real-time timeslot assignment algorithms for ISM networks.

Optimal prioritized channel allocation in cellular mobile systems

Abstract Under the cutoff priority discipline, the optimal prioritized channel allocation problem is formulated, which minimizes the weighted average blocking probability of handoff calls while ensuring the prespecified grade of service for new calls and the co-channel interference constraints. We use the concept of pattern to deal with the problem more conveniently. Using Lagrangean relaxation and subgradient optimization techniques, we obtain high-quality solutions with information about their deviations from true optimal solutions. Computational experiments show that our method works very well. Scope and Purpose Channel allocation is one of the most important problems in the design of cellular mobile systems. Since the number of cells of forthcoming networks will increase rapidly, this problem will be of even greater importance in the future. In cellular mobile systems, a new channel should be assigned by the new base when a call enters an adjacent cell. It provides continuation of ongoing calls as the user travels across cell boundaries, and is called handoff. The handoff call is forced to terminate before completion if there are no channels available in the new cell. However, in many practical situations, the blocking of a handoff call attempt is critical since it will result in a disconnection of the call in the middle of conversation. Thus, for reducing the blocking probability of handoff calls, several algorithms based on the cutoff priority scheme have been introduced in the literature. In the cutoff priority scheme, priority is given to handoff calls by exclusively reserving some channels called guard channels for them. In this article, we formulate a prioritized channel allocation problem under the cutoff priority scheme in general multicell environments, and suggest an efficient algorithm to solve that problem.

Channel allocation in cellular radio networks

In this article, two efficient heuristic algorithms are suggested for the channel allocation problem which minimizes the average blocking probability of the whole network subject to the co-channel, adjacent-site and co-site interference constraints, given the number of available channels. We convert this problem into a convenient form using the piecewise linearization technique and the concept of pattern, and apply Lagrangean relaxation and subgradient optimization techniques. Computational experiments show that this procedure provides high-quality solutions with information about their error ranges for networks with special compatibility matrices. We also suggest a general procedure using a GOS (grade of service) updating scheme, and provide encouraging computational results.

Optimal timeslot and channel allocation considering fairness for multicell CDMA/TDD systems

The CDMA/TDD system is a highly attractive solution to support the next generation cellular mobile systems which provide unbalanced multimedia services between downlink and uplink. In this paper, we analyze the interference for downlink and uplink timeslots in a multicell CDMA/TDD system. We also mathematically formulate an optimal timeslot and channel allocation problem considering capacity fairness among cells, which is to maximize the system capacity under the given traffic unbalance, and propose an efficient algorithm based on the simulated annealing technique. Extensive experimental results show that the proposed scheme yields a good performance, and fairness among cells improves with a decrease in the system capacity.

An efficient local predictive method for distributed timeslot allocation in CDMA/TDD

Slot allocation in a CDMA/TDD cellular network, which decomposes a frame into uplink and downlink slots, is a multi-dimensional optimization problem involving all the cells in the system due to interference between cells. However, the complexity of the problem makes it computationally too expensive to solve simultaneously for all the cells in a system of reasonable size. In this paper, we propose a novel predictive and distributed slot allocation scheme for transporting multimedia traffic over CDMA/TDD cellular networks with multiple cells, designed to maximize a general utility function that is (strictly) concave increasing. We formulate the utility-maximizing slot allocation problem as a nonlinear integer programming problem, and provide a solution that can be realized efficiently in a distributed manner. Also, we derive sufficient conditions for a real-number solution to be optimal and develop an efficient algorithm to find a near-optimal integer solution. The effectiveness of the integer solution is demonstrated by showing that its gap with the optimal solution is small. We show by experiments that the proposed predictive method has a sufficient accuracy when there are many active users in cells. Extensive simulation results show that the proposed distributed slot allocation method performs very well in comparison to an existing method.

Capacity of multicell CDMA/shared-TDD system

The CDMA/shared-TDD system is a highly attractive solution to support the next generation cellular mobile systems which provide unbalanced multimedia services. In this paper, we analyze the interference for downlink, uplink, and crossed timeslots in a multicell CDMA/shared-TDD system. We also mathematically formulate optimal timeslots and channel allocation problem, which is to maximize the system capacity under the given traffic unbalance. Computational experiments show that the capacity of the CDMA/shared-TDD system is larger than that of the pure CDMA/TDD system.

Information Exchange Based Low-Complexity Slot Allocation in TDD-CDMA Cellular Systems

In this paper, we consider a predictive and distributed slot allocation scheme for transporting multi-class traffic over TDD-CDMA cellular networks with multiple cells, designed to maximize a general utility function that is concave increasing. We formulate the utility-maximizing slot allocation problem as a nonlinear integer programming problem, and provide a solution that can be realized efficiently in a distributed manner. Also, we derive sufficient conditions for a real-number solution to be optimal and develop an efficient algorithm to find a near-optimal integer solution. The effectiveness of the integer solution is demonstrated by showing that its gap with the optimal solution is small. We show by experiments that the proposed predictive method has a sufficient accuracy when there are many active users in cells. Extensive simulation results show that the proposed distributed slot allocation method performs very well in comparison to an existing method.