• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.29 no.1D
• /
• pp.25-32
• /
• 2009

Currently, various types of TDM (Transportation Demand Management) policies are being studied and implemented in an attempt to overcome the limitations of supply oriented policies. In this context, this paper addressed issues of effectiveness and possible domestic implementation of the HOT lane. The possible site of implementation selected for this simulation study is part of the Kyung-bu freeway, where a dedicated bus lane is currently being operated. Minimum length of distance required in between interchanges and access points of the HOT lane for vehicles to safely enter and exit the lane, and traffic management policies for effectively managing the weaving traffic trying to enter and exit the HOT lane were presented. A 5.2km section of freeway from Ki-heuing IC to Suwon IC and a 8.3km section from Hak-uei JC to Pan-gyo JC have been selected as possible sites of implementation for the HOT lane, in which congestion occurs regularly due to the high level of travel demand. VISSIM simulation program has been used to analyze the effects of the HOT lane under the assumption that one-lane HOT lane has been put into operation in these sections and that the lane change rate were in between 5% to 30%. The results of each possible scenario have proven that overall travel speed on the general lanes have increased as well by 1.57~2.62km/h after the implementation of the HOT lane. It is meaningful that this study could serve as a basic reference data for possible follow-up studies on the HOT lane as one effective method of TDM policies. Considering that the bus travel rate would continue increase and assuming the improvement in travel speed on general lanes, similar case study can be implemented where gaps between buses on bus lane are available, as a possible alternative of efficient bus lane management policies.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.71-88
• /
• 2017

Language models were originally developed for speech recognition and language processing. Using a set of example sentences, a language model predicts the next word or character based on sequential input data. N-gram models have been widely used but this model cannot model the correlation between the input units efficiently since it is a probabilistic model which are based on the frequency of each unit in the training set. Recently, as the deep learning algorithm has been developed, a recurrent neural network (RNN) model and a long short-term memory (LSTM) model have been widely used for the neural language model (Ahn, 2016; Kim et al., 2016; Lee et al., 2016). These models can reflect dependency between the objects that are entered sequentially into the model (Gers and Schmidhuber, 2001; Mikolov et al., 2010; Sundermeyer et al., 2012). In order to learning the neural language model, texts need to be decomposed into words or morphemes. Since, however, a training set of sentences includes a huge number of words or morphemes in general, the size of dictionary is very large and so it increases model complexity. In addition, word-level or morpheme-level models are able to generate vocabularies only which are contained in the training set. Furthermore, with highly morphological languages such as Turkish, Hungarian, Russian, Finnish or Korean, morpheme analyzers have more chance to cause errors in decomposition process (Lankinen et al., 2016). Therefore, this paper proposes a phoneme-level language model for Korean language based on LSTM models. A phoneme such as a vowel or a consonant is the smallest unit that comprises Korean texts. We construct the language model using three or four LSTM layers. Each model was trained using Stochastic Gradient Algorithm and more advanced optimization algorithms such as Adagrad, RMSprop, Adadelta, Adam, Adamax, and Nadam. Simulation study was done with Old Testament texts using a deep learning package Keras based the Theano. After pre-processing the texts, the dataset included 74 of unique characters including vowels, consonants, and punctuation marks. Then we constructed an input vector with 20 consecutive characters and an output with a following 21st character. Finally, total 1,023,411 sets of input-output vectors were included in the dataset and we divided them into training, validation, testsets with proportion 70:15:15. All the simulation were conducted on a system equipped with an Intel Xeon CPU (16 cores) and a NVIDIA GeForce GTX 1080 GPU. We compared the loss function evaluated for the validation set, the perplexity evaluated for the test set, and the time to be taken for training each model. As a result, all the optimization algorithms but the stochastic gradient algorithm showed similar validation loss and perplexity, which are clearly superior to those of the stochastic gradient algorithm. The stochastic gradient algorithm took the longest time to be trained for both 3- and 4-LSTM models. On average, the 4-LSTM layer model took 69% longer training time than the 3-LSTM layer model. However, the validation loss and perplexity were not improved significantly or became even worse for specific conditions. On the other hand, when comparing the automatically generated sentences, the 4-LSTM layer model tended to generate the sentences which are closer to the natural language than the 3-LSTM model. Although there were slight differences in the completeness of the generated sentences between the models, the sentence generation performance was quite satisfactory in any simulation conditions: they generated only legitimate Korean letters and the use of postposition and the conjugation of verbs were almost perfect in the sense of grammar. The results of this study are expected to be widely used for the processing of Korean language in the field of language processing and speech recognition, which are the basis of artificial intelligence systems.

• Journal of Intelligence and Information Systems
• /
• v.27 no.1
• /
• pp.191-207
• /
• 2021

Up to this day, mobile communications have evolved rapidly over the decades, mainly focusing on speed-up to meet the growing data demands of 2G to 5G. And with the start of the 5G era, efforts are being made to provide such various services to customers, as IoT, V2X, robots, artificial intelligence, augmented virtual reality, and smart cities, which are expected to change the environment of our lives and industries as a whole. In a bid to provide those services, on top of high speed data, reduced latency and reliability are critical for real-time services. Thus, 5G has paved the way for service delivery through maximum speed of 20Gbps, a delay of 1ms, and a connecting device of 106/㎢ In particular, in intelligent traffic control systems and services using various vehicle-based Vehicle to X (V2X), such as traffic control, in addition to high-speed data speed, reduction of delay and reliability for real-time services are very important. 5G communication uses high frequencies of 3.5Ghz and 28Ghz. These high-frequency waves can go with high-speed thanks to their straightness while their short wavelength and small diffraction angle limit their reach to distance and prevent them from penetrating walls, causing restrictions on their use indoors. Therefore, under existing networks it's difficult to overcome these constraints. The underlying centralized SDN also has a limited capability in offering delay-sensitive services because communication with many nodes creates overload in its processing. Basically, SDN, which means a structure that separates signals from the control plane from packets in the data plane, requires control of the delay-related tree structure available in the event of an emergency during autonomous driving. In these scenarios, the network architecture that handles in-vehicle information is a major variable of delay. Since SDNs in general centralized structures are difficult to meet the desired delay level, studies on the optimal size of SDNs for information processing should be conducted. Thus, SDNs need to be separated on a certain scale and construct a new type of network, which can efficiently respond to dynamically changing traffic and provide high-quality, flexible services. Moreover, the structure of these networks is closely related to ultra-low latency, high confidence, and hyper-connectivity and should be based on a new form of split SDN rather than an existing centralized SDN structure, even in the case of the worst condition. And in these SDN structural networks, where automobiles pass through small 5G cells very quickly, the information change cycle, round trip delay (RTD), and the data processing time of SDN are highly correlated with the delay. Of these, RDT is not a significant factor because it has sufficient speed and less than 1 ms of delay, but the information change cycle and data processing time of SDN are factors that greatly affect the delay. Especially, in an emergency of self-driving environment linked to an ITS(Intelligent Traffic System) that requires low latency and high reliability, information should be transmitted and processed very quickly. That is a case in point where delay plays a very sensitive role. In this paper, we study the SDN architecture in emergencies during autonomous driving and conduct analysis through simulation of the correlation with the cell layer in which the vehicle should request relevant information according to the information flow. For simulation: As the Data Rate of 5G is high enough, we can assume the information for neighbor vehicle support to the car without errors. Furthermore, we assumed 5G small cells within 50 ~ 250 m in cell radius, and the maximum speed of the vehicle was considered as a 30km ~ 200 km/hour in order to examine the network architecture to minimize the delay.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.1
• /
• pp.51-60
• /
• 2014

In this study, ICT medical service provider's level of knowledge fire fighting safety and methods on coping with fires in the regions of Gwangju and Jeonam Province of Korea were investigated to determine the elements affecting such levels and provide basic information on the manuals for educating how to cope with the fire fighting safety in medical facilities. The data were analyzed using SPSS Win 14.0. The scores of level of knowledge fire fighting safety of ICT medical service provider's were 7.06(10 point scale), and the scores of level of recognizing how to cope with fire fighting safety were 6.61(11 point scale). level of recognizing how to cope with fire fighting safety were significantly different according to gender(t=4.12, p<.001), age(${\chi}^2$=17.24, p<.001), length of career(${\chi}^2$=22.76, p<.001), experience with fire fighting safety education(t=6.10, p<.001), level of subjective knowledge on fire fighting safety(${\chi}^2$=53.83, p<.001). In order to enhance the level of understanding of fire fighting safety and methods of coping by the ICT medical service providers it is found that: self-directed learning through avoiding the education just conveying knowledge by lecture tailored learning for individuals fire fighting education focused on experiencing actual work by developing various contents emphasizing cooperative learning deploying patients by classification systems using simulations and a study on the implementation of digital anti-fire monitoring system with multipoint communication protocol, a design and development of the smoke detection system using infra-red laser for fire detection in the wide space, video based fire detection algorithm using gaussian mixture mode developing an education manual for coping with fire fighting safety through multi learning approach at the medical facilities are required.

• Nuclear Medicine and Molecular Imaging
• /
• v.42 no.6
• /
• pp.469-477
• /
• 2008

Purpose: The goal of this paper is to present the design and performance of a position encoding circuit for $16{\times}16$ array of position sensitive multi-anode photomultiplier tube for small animal PET scanners. This circuit which reduces the number of readout channels from 256 to 4 channels is based on a charge division method utilizing a resistor array. Materials and Methods: The position encoding circuit was simulated with PSpice before fabrication. The position encoding circuit reads out the signals from H9500 flat panel PMTs (Hamamatsu Photonics K.K., Japan) on which $1.5{\times}1.5{\times}7.0\;mm^3$ $L_{0.9}GSO$ ($Lu_{1.8}Gd_{0.2}SiO_{5}:Ce$) crystals were mounted. For coincidence detection, two different PET modules were used. One PET module consisted of a $29{\times}29\;L_{0.9}GSO$ crystal layer, and the other PET module two $28{\times}28$ and $29{\times}29\;L_{0.9}GSO$ crystal layers which have relative offsets by half a crystal pitch in x- and y-directions. The crystal mapping algorithm was also developed to identify crystals. Results: Each crystal was clearly visible in flood images. The crystal identification capability was enhanced further by changing the values of resistors near the edge of the resistor array. Energy resolutions of individual crystal were about 11.6%(SD 1.6). The flood images were segmented well with the proposed crystal mapping algorithm. Conclusion: The position encoding circuit resulted in a clear separation of crystals and sufficient energy resolutions with H9500 flat-panel PMT and $L_{0.9}GSO$ crystals. This circuit is good enough for use in small animal PET scanners.

• Journal of Biosystems Engineering
• /
• v.4 no.2
• /
• pp.65-83
• /
• 1979

Low-temperature drying systems have been extensively used for drying cereal grain such as shelled corn and wheat. Since the 1973 energy crisis, many researches have been conducted to apply solar energy as supplemental heat to natural air drying systems. However, little research on rough rice drying has been done in this area, especially very little in Korea. In designing a solar drying system, quality loss, airflow requirements, temperature rise of drying air, fan power and energy requirements should be throughly studied. The factors affecting solar drying systems are airflow rate, initial moisture content, the amount of heat added to drying air, fan operation method and the weather conditions. The major objectives of this study were to analyze the effects of the performance factors and determine design parameters such as airflow requirements, optimum bed depth, optimum temperature rise of drying air, fan operation method and collector size. Three hourly observations based on the 4-year weather data in Chuncheon area were used to simulate rough rice drying. The results can be summarized as follows: 1. The results of the statistical analysis indicated that the experimental and predicted values of the temperature rise of the air passing through the collector agreed well. 2. Equilibrium moisture content was affected a little by airflow rate, but affected mainly by the amount of heat added, to drying air. Equilibrium moisture content ranged from 12.2 to 13.2 percent wet basis for the continuous fan operation, from 10.4 to 11.7 percent wet basis for the intermittent fan operation respectively, in range of 1. 6 to 5. 9 degrees Centigrade average temperature rise of drying air. 3. Average moisture content when top layer was dried to 15 percent wet basis ranged from 13.1 to 13.9 percent wet basis for the continuous fan operation, from 11.9 to 13.4 percent wet basis for the intermittent fan operation respectively, in the range of 1.6 to 5.9 degrees Centigrade average temperature rise of drying air and 18 to 24 percent wet basis initial moisture content. The results indicated that grain was overdried with the intermittent fan operation in any range of temperature rise of drying air. Therefore, the continuous fan operation is usually more effective than the intermittent fan operation considering the overdrying. 4. For the continuous fan operation, the average temperature rise of drying air may be limited to 2.2 to 3. 3 degrees Centigrade considering safe storage moisture level of 13.5 to 14 perceut wet basis. 5. Required drying time decrease ranged from 40 to 50 percent each time the airflow rate was doubled and from 3.9 to 4.3 percent approximately for each one degrees Centigrade in average temperature rise of drying air regardless of the fan operation methods. Therefore, the average temperature rise of drying air had a little effect on required drying time. 6. Required drying time increase ranged from 18 to 30 percent approximately for each 2 percent increase in initial moisture content regardless of the fan operation methods, in the range of 18 to 24 percent moisture. 7. The intermittent fan operation showed about 36 to 42 percent decrease in required drying time as compared with the continuous fan operation. 8. Drymatter loss decrease ranged from 34 to 46 percent each time the airflow rate was doubled and from 2 to 3 percent approximately for each one degrees Centigrade in average temperature rise of drying air, regardless of the fan operation methods. Therefore, the average temperature rise of drying air had a little effect on drymatter loss. 9. Drymatter loss increase ranged from 50 to 78 percent approximately for each 2 percent increase in initial moisture content, in the range of 18 to 24 percent moisture. 10. The intermittent fan operation: showed about 40 to 50 percent increase in drymatter loss as compared with the continuous fan operation and the increasing rate was higher at high level of initial moisture and average temperature rise. 11. Year-to-year weather conditions had a little effect on required drying time and drymatter loss. 12. The equations for estimating time required to dry top layer to 16 and 1536 wet basis and drymatter loss were derived as functions of the performance factors. by the least square method. 13. Minimum airflow rates based on 0.5 percent drymatter loss were estimated. Minimum airflow rates for the intermittent fan operation were approximately 1.5 to 1.8 times as much as compared with the continuous fan operation, but a few differences among year-to-year. 14. Required fan horsepower and energy for the intermittent fan operation were 3. 7 and 1. 5 times respectively as much as compared with the continuous fan operation. 15. The continuous fan operation may be more effective than the intermittent fan operation considering overdrying, fan horsepower requirements, and energy use. 16. A method for estimating the required collection area of flat-plate solar collector using average temperature rise and airflow rate was presented.

• Journal of Biosystems Engineering
• /
• v.4 no.2
• /
• pp.64-64
• /
• 1979

Low-temperature drying systems have been extensively used for drying cereal grain such as shelled corn and wheat. Since the 1973 energy crisis, many researches have been conducted to apply solar energy as supplemental heat to natural air drying systems. However, little research on rough rice drying has been done in this area, especially very little in Korea. In designing a solar drying system, quality loss, airflow requirements, temperature rise of drying air, fan power and energy requirements should be throughly studied. The factors affecting solar drying systems are airflow rate, initial moisture content, the amount of heat added to drying air, fan operation method and the weather conditions. The major objectives of this study were to analyze the effects of the performance factors and determine design parameters such as airflow requirements, optimum bed depth, optimum temperature rise of drying air, fan operation method and collector size. Three hourly observations based on the 4-year weather data in Chuncheon area were used to simulate rough rice drying. The results can be summarized as follows: 1. The results of the statistical analysis indicated that the experimental and predicted values of the temperature rise of the air passing through the collector agreed well.2. Equilibrium moisture content was affected a little by airflow rate, but affected mainly by the amount of heat added, to drying air. Equilibrium moisture content ranged from 12.2 to 13.2 percent wet basis for the continuous fan operation, from 10.4 to 11.7 percent wet basis for the intermittent fan operation respectively, in range of 1. 6 to 5. 9 degrees Centigrade average temperature rise of drying air.3. Average moisture content when top layer was dried to 15 percent wet basis ranged from 13.1 to 13.9 percent wet basis for the continuous fan operation, from 11.9 to 13.4 percent wet basis for the intermittent fan operation respectively, in the range of 1.6 to 5.9 degrees Centigrade average temperature rise of drying air and 18 to 24 percent wet basis initial moisture content. The results indicated that grain was overdried with the intermittent fan operation in any range of temperature rise of drying air. Therefore, the continuous fan operation is usually more effective than the intermittent fan operation considering the overdrying.4. For the continuous fan operation, the average temperature rise of drying air may be limited to 2.2 to 3. 3 degrees Centigrade considering safe storage moisture level of 13.5 to 14 perceut wet basis.5. Required drying time decrease ranged from 40 to 50 percent each time the airflow rate was doubled and from 3.9 to 4.3 percent approximately for each one degrees Centigrade in average temperature rise of drying air regardless of the fan operation methods. Therefore, the average temperature rise of drying air had a little effect on required drying time.6. Required drying time increase ranged from 18 to 30 percent approximately for each 2 percent increase in initial moisture content regardless of the fan operation methods, in the range of 18 to 24 percent moisture.7. The intermittent fan operation showed about 36 to 42 percent decrease in required drying time as compared with the continuous fan operation.8. Drymatter loss decrease ranged from 34 to 46 percent each time the airflow rate was doubled and from 2 to 3 percent approximately for each one degrees Centigrade in average temperature rise of drying air, regardless of the fan operation methods. Therefore, the average temperature rise of drying air had a little effect on drymatter loss. 9. Drymatter loss increase ranged from 50 to 78 percent approximately for each 2 percent increase in initial moisture content, in the range of 18 to 24 percent moisture. 10. The intermittent fan operation: showed about 40 to 50 percent increase in drymatter loss as compared with the continuous fan operation and the increasing rate was higher at high level of initial moisture and average temperature rise.11. Year-to-year weather conditions had a little effect on required drying time and drymatter loss.12. The equations for estimating time required to dry top layer to 16 and 1536 wet basis and drymatter loss were derived as functions of the performance factors. by the least square method.13. Minimum airflow rates based on 0.5 percent drymatter loss were estimated.Minimum airflow rates for the intermittent fan operation were approximately 1.5 to 1.8 times as much as compared with the continuous fan operation, but a few differences among year-to-year.14. Required fan horsepower and energy for the intermittent fan operation were3. 7 and 1. 5 times respectively as much as compared with the continuous fan operation.15. The continuous fan operation may be more effective than the intermittent fan operation considering overdrying, fan horsepower requirements, and energy use.16. A method for estimating the required collection area of flat-plate solar collector using average temperature rise and airflow rate was presented.

• KIPS Transactions on Computer and Communication Systems
• /
• v.6 no.1
• /
• pp.1-8
• /
• 2017

Communications for mobile sink groups such as rescue teams or platoons bring about a new challenging issue for handling mobility in wireless sensor networks. To do this, many studies have been proposed to support mobile sink groups. When closely looking at mobile sink groups, they can be divided into (multiple) small groups according to the property of applications. For example, a platoon can be divided into multiple squads to carry out its mission in the battle field. However, the previous studies cannot efficiently support the division of mobile sink groups because they do not address three challenging issues engendered by the mobile sink group division. The first issue is to select a leader sink for a new small mobile sink group. The efficient data delivery from a source to small mobile sink groups is the second issue. Last, the third issue is to share data between leader sinks of small mobile sink groups. Thus, this paper proposes a routing protocol to efficiently support the division of mobile sink groups by solving the three challenging issues. For the first issue, the proposed protocol selects a leader sink of a new small mobile sink group which provide a minimum summation of the distance between the new leader sink and the previous leader sink and the distance from the new leader sink to all of its member sinks. For the efficient data delivery from a source to small mobile sink groups in the second issue, the proposed protocol determines the path to minimize the data dissemination distance from source to small mobile sink group by calculating with the location information of both the source and the leader sinks. With regard to the third issue, the proposed protocol exploits member sinks located among leader sinks to provide efficient data sharing among leaders sinks by considering the location information of member sinks. Simulation results verified that the proposed protocol is superior to the previous protocol in terms of the energy consumption.

• Progress in Medical Physics
• /
• v.23 no.3
• /
• pp.145-153
• /
• 2012

In this study, we quantify the residual motion artifact in 4D-CT scan using the dynamic lung phantom which could simulate respiratory target motion and suggest a simple one-dimension theoretical model to explain and characterize the source of motion artifacts in 4DCT scanning. We set-up regular 1D sine motion and adjusted three level of amplitude (10, 20, 30 mm) with fixed period (4s). The 4DCT scans are acquired in helical mode and phase information provided by the belt type respiratory monitoring system. The images were sorted into ten phase bins ranging from 0% to 90%. The reconstructed images were subsequently imported into the Treatment Planning System (CorePLAN, SC&J) for target delineation using a fixed contour window and dimensions of the three targets are measured along the direction of motion. Target dimension of each phase image have same changing trend. The error is minimum at 50% phase in all case (10, 20, 30 mm) and we found that ${\Delta}S$ (target dimension change) of 10, 20 and 30 mm amplitude were 0 (0%), 0.1 (5%), 0.1 (5%) cm respectively compare to the static image of target diameter (2 cm). while the error is maximum at 30% and 80% phase ${\Delta}S$ of 10, 20 and 30 mm amplitude were 0.2 (10%), 0.7 (35%), 0.9 (45%) cm respectively. Based on these result, we try to analysis the residual motion artifact in 4D-CT scan using a simple one-dimension theoretical model and also we developed a simulation program. Our results explain the effect of residual motion on each phase target displacement and also shown that residual motion artifact was affected that the target velocity at each phase. In this study, we focus on provides a more intuitive understanding about the residual motion artifact and try to explain the relationship motion parameters of the scanner, treatment couch and tumor. In conclusion, our results could help to decide the appropriate reconstruction phase and CT parameters which reduce the residual motion artifact in 4DCT.

• Journal of Intelligence and Information Systems
• /
• v.18 no.4
• /
• pp.43-57
• /
• 2012

To enhance the competitive advantage in a constantly changing business environment, an enterprise management must make the right decision in many business activities based on both internal and external information. Thus, providing accurate information plays a prominent role in management's decision making. Intuitively, historical data can provide a feasible estimate through the forecasting models. Therefore, if the service department can estimate the service quantity for the next period, the service department can then effectively control the inventory of service related resources such as human, parts, and other facilities. In addition, the production department can make load map for improving its product quality. Therefore, obtaining an accurate service forecast most likely appears to be critical to manufacturing companies. Numerous investigations addressing this problem have generally employed statistical methods, such as regression or autoregressive and moving average simulation. However, these methods are only efficient for data with are seasonal or cyclical. If the data are influenced by the special characteristics of product, they are not feasible. In our research, we propose a forecasting framework that predicts service demand of manufacturing organization by combining Case-based reasoning (CBR) and leveraging an unsupervised artificial neural network based clustering analysis (i.e., Self-Organizing Maps; SOM). We believe that this is one of the first attempts at applying unsupervised artificial neural network-based machine-learning techniques in the service forecasting domain. Our proposed approach has several appealing features : (1) We applied CBR and SOM in a new forecasting domain such as service demand forecasting. (2) We proposed our combined approach between CBR and SOM in order to overcome limitations of traditional statistical forecasting methods and We have developed a service forecasting tool based on the proposed approach using an unsupervised artificial neural network and Case-based reasoning. In this research, we conducted an empirical study on a real digital TV manufacturer (i.e., Company A). In addition, we have empirically evaluated the proposed approach and tool using real sales and service related data from digital TV manufacturer. In our empirical experiments, we intend to explore the performance of our proposed service forecasting framework when compared to the performances predicted by other two service forecasting methods; one is traditional CBR based forecasting model and the other is the existing service forecasting model used by Company A. We ran each service forecasting 144 times; each time, input data were randomly sampled for each service forecasting framework. To evaluate accuracy of forecasting results, we used Mean Absolute Percentage Error (MAPE) as primary performance measure in our experiments. We conducted one-way ANOVA test with the 144 measurements of MAPE for three different service forecasting approaches. For example, the F-ratio of MAPE for three different service forecasting approaches is 67.25 and the p-value is 0.000. This means that the difference between the MAPE of the three different service forecasting approaches is significant at the level of 0.000. Since there is a significant difference among the different service forecasting approaches, we conducted Tukey's HSD post hoc test to determine exactly which means of MAPE are significantly different from which other ones. In terms of MAPE, Tukey's HSD post hoc test grouped the three different service forecasting approaches into three different subsets in the following order: our proposed approach > traditional CBR-based service forecasting approach > the existing forecasting approach used by Company A. Consequently, our empirical experiments show that our proposed approach outperformed the traditional CBR based forecasting model and the existing service forecasting model used by Company A. The rest of this paper is organized as follows. Section 2 provides some research background information such as summary of CBR and SOM. Section 3 presents a hybrid service forecasting framework based on Case-based Reasoning and Self-Organizing Maps, while the empirical evaluation results are summarized in Section 4. Conclusion and future research directions are finally discussed in Section 5.