A key task of managing an inbound call center is in estimating its performance and consequently plan its capacity, which can be considered a complex task since several system variables are stochastic. These issues are highly crucial for certain time-sensitive services, such as emergency call services. Waiting times affect the service quality of call centers in general, but various customers may place different waiting time expectancies depending on the need. Call center managers struggle to find the relationship between these expectations to their strategical, tactical and operational issues. They are assisted by queueing models that approximate the outcome. Simple setups use analytical approximations while a network of multi-skilled agents serving several customer classes is dependent on computer simulations. Regardless of simple or complex setups, models assume that the system components are homogenous, that the components have some parametric distribution, and that they remain the same regardless of the setup. Human resource and marketing research show that such status quo assumptions are not highly reliable. As an example, customer experience is often affected by the skill of the agent, and agents themselves are affected by their workload and duties, which inter alia affect their efficiency. This thesis aim to assist the Swedish emergency call center with a strategical issue, which require detection of some causalities in the set of system components. The overall aim is to design a simulation model, but such model requires a lot of detailed system knowledge, which itself adds to the knowledge gap in the research field. Findings that contribute to the scientific knowledge body include the burst model that addresses some of the non-stationarity of call arrivals, since some rapid rate increments derives from a latent emergency event. Other contributions are the introduction of stochastic agent behavior, which increases the uncertainty in queueing models; and the service time relationship to geographical distance. The latter may involve general evidence on how area-specific understanding and cultural differences affect the quality of service. This is important for organizations that consider off-shoring or outsourcing their call center service. These findings, along with several undiscovered and unknown influencers, are needed in order to design a reliable simulation model. However, the proposed model in this study cannot be rejected, in terms of waiting time replication. This robust model allowed traffic routing strategies to be evaluated and also assisted managers of the emergency call center into a strategical shift in the late 2015. / <p>Vid tidpunkten för framläggningen av avhandlingen var följande delarbeten opublicerade: delarbete 1 och 3 inskickat.</p><p>At the time of the defence the following papers were unpublished: paper 1 and 3 submitted.</p>

This work describes the rationalization process in Lotus Notes in a company that provides complete support for Lotus Notes Product for a large international company. It describes three level support and also deals with the rationalization process between the first level of support (customer service) and second level support (Lotus Notes administration). Finally, describes how to create – programming the agents in the Lotus Notes team for the second level, which will help to speed up and improve the services provided by this team.

Time series is a collection of points gathered at regular intervals. Time series analysis explores the time correlations and tries to model it according to trend and seasonality. One of the most relevant tasks, in time series analysis, is forecasting future values, which is considered fundamental in many real-world scenarios. Nowadays, many companies forecast using hand-written models or naive statistical models. Call centers are the front end of the organization, managing the relationship with the customers. A key challenge for call centers remains the call load forecast and the optimization of the schedule. Call load indicates the number of calls a call center receives. The call load forecast is mostly exploited to schedule the staff. They are interested in the short term forecast to handle the unforeseen and to optimize the staff schedule, and in the long term forecast to hire or assign staff to other tasks. Machine learning has been applied to several fields reporting excellent results, and recently, time series forecasting problems have gained a high-interest thanks to the new recurrent network, named Long-short Term Memory. This thesis has explored the capabilities of machine learning in modeling and forecasting call load time series, characterized by a strong seasonality, both at daily and hourly scale. We compare Seasonal Artificial Neural Network (ANN) and a Long-Short Term Memory (LSTM) models with Seasonal Autoregressive Integrated Moving Average (SARIMA) model, which is one of the most common statistical method utilized by call centers. The primary metric used to evaluate the results is the Normalized Mean Squared Error (NMSE), the secondary is the Symmetric Mean Absolute Percentage Error (SMAPE), utilized to calculate the accuracy of the models. We carried out our experiments on three different datasets provided by the Teleopti. Experimental results have proven SARIMA to be more accurate in forecasting at daily scale across the three datasets. It performs better than the Seasonal ANN and the LSTM with a limited amount of data points. At hourly scale, Seasonal ANN and LSTM outperform SARIMA, showing robustness across a forecasting horizon of 160 points. Finally, SARIMA has shown no correlation between the quality of the model and the number of data points, while both SANN and LSTM improves together with the number of sample / Tidsserie är en samling punkter som samlas in med jämna mellanrum. Tidsseriens analys undersöker tidskorrelationerna och försöker modellera den enligt trend och säsongsbetonade. En av de mest relevanta uppgifterna, i tidsserieranalys, är att förutse framtida värden, som anses vara grundläggande i många verkliga scenarier. Numera förutspår många företag med handskrivna modeller eller naiva statistiska modeller. Callcenter är organisationens främre del och hanterar relationen med kunderna. En viktig utmaning för callcentra är fortfarande samtalslastprognosen och optimeringen av schemat. Samtalslast indikerar antalet samtal ett callcenter tar emot. Samtalslastprognosen utnyttjas mest för att schemalägga personalen. De är intresserade av den kortsiktiga prognosen för att hantera det oförutsedda och för att optimera personalplanen och på långsiktigt prognos för att anställa eller tilldela personal till andra uppgifter. Maskininlärning har använts på flera fält som rapporterar utmärkta resultat, och nyligen har prognosproblem i tidsserier fått ett stort intresse tack vare det nya återkommande nätverket, som heter Long-short Term Memory. Den här avhandlingen har undersökt kapaciteten för maskininlärning i modellering och prognoser samtalsbelastningstidsserier, kännetecknad av en stark säsongsbetonning, både på daglig och timskala. Vi jämför modeller med säsongsmässigt artificiellt neuralt nätverk (ANN) och ett LSTM-modell (Long- Short Term Memory) med Seasonal Autoregressive Integrated Moving Average (SARIMA)-modell, som är en av de vanligaste statistiska metoderna som används av callcenter. Den primära metriken som används för att utvärdera resultaten är det normaliserade medelkvadratfelet (NMSE), det sekundära är det symmetriska genomsnittet absolut procentuellt fel (SMAPE), som används för att beräkna modellernas noggrannhet. Vi genomförde våra experiment på tre olika datasätt från Teleopti. Experimentella resultat har visat att SARIMA är mer exakt när det gäller prognoser i daglig skala över de tre datasätten. Det presterar bättre än Seasonal ANN och LSTM med en begränsad mängd datapoäng. På timskala överträffar Seasonal ANN och LSTM SARIMA och visar robusthet över en prognoshorisont på 160 poäng. SARIMA har slutligen inte visat någon korrelation mellan modellens kvalitet och antalet datapunkter, medan både SANN och LSTM förbättras tillsammans med antalet sampel.

The objective of this study was to: (1) assess the overall wellness climate profile of a sample of higher education call centre employees for national benchmarking purposes; (2) explore the relationship between the participants‟ sense of coherence, affective wellbeing and burnout; and (3) determine how the participants differ regarding these variables in terms of socio-demographic contextual factors such as gender, race, age, and marital status. The South African Employee Health and Wellness Survey was used as a measuring instrument. Compared to the national norm, the results indicated a risky wellness climate reflecting a burnout propensity, lower morale (affective wellbeing) and lower resilience (sense of coherence). Significant relations existed between the participants‟ sense of coherence, affective wellbeing and burnout levels. Significant differences regarding these variables were also detected between males and females and the various marital status groups regarding the participants‟ sense of coherence, affective wellbeing and burnout. The findings of this study contributed new knowledge that may be used to inform employee wellness programmes within a higher education call centre environment. The study concluded with recommendations for future research and practice. / Industrial and Organisational Psychology / M. Comm. (Industrial and Organisational Psychology)

The objective of this study was to: (1) assess the overall wellness climate profile of a sample of higher education call centre employees for national benchmarking purposes; (2) explore the relationship between the participants‟ sense of coherence, affective wellbeing and burnout; and (3) determine how the participants differ regarding these variables in terms of socio-demographic contextual factors such as gender, race, age, and marital status. The South African Employee Health and Wellness Survey was used as a measuring instrument. Compared to the national norm, the results indicated a risky wellness climate reflecting a burnout propensity, lower morale (affective wellbeing) and lower resilience (sense of coherence). Significant relations existed between the participants‟ sense of coherence, affective wellbeing and burnout levels. Significant differences regarding these variables were also detected between males and females and the various marital status groups regarding the participants‟ sense of coherence, affective wellbeing and burnout. The findings of this study contributed new knowledge that may be used to inform employee wellness programmes within a higher education call centre environment. The study concluded with recommendations for future research and practice. / Industrial and Organisational Psychology / M. Comm. (Industrial and Organisational Psychology)

<p>I Datainspektionens rapporter 2003:3 och 2005:3 har myndigheten granskat hur personuppgifter behandlas i samband med arbetsgivares övervakning av anställda. Resultaten av dessa rapporter har visat på förekommande brister i förhållande till personuppgiftslagen i många av de granskade verksamheterna.</p><p>Mot denna bakgrund har en fallstudie av ett medelstort callcenterföretag genomförts för att ur ett tekniskt perspektiv undersöka hur väl personuppgiftslagen följs i samband med företagets övervakning av sin personal. Utifrån brister identifierade i fallstudien har en mindre kartläggning av liknande företag genomförts i ett försök att mäta graden av generaliserbarhet i de identifierade bristerna.</p><p>De brister som har identifierats i fallstudien rör i första hand otillfredsställande rutiner för gallring av personuppgifter vid e-postövervakning samt avsaknad av information till de anställda om vilka kontroller som sker i samband med detta. Behov av en uppdatering av företagets IT-policy samt ett behov av en övervakningslösning för personalens internetanvändande kunde också noteras.</p><p>Kartläggningen har givit antydningar om att liknande brister vad gäller rutiner för gallring samt bristande information kan tänkas förekomma även hos andra medelstora callcenterföretag.</p><p>Slutsatsen av arbetet är att flera av de brister som uppdagades av Datainspektionen för fem år sedan tycks göra sig gällande bland vissa medelstora callcenterföretag även idag.</p> / <p>In the 2003:3 and 2005:3 reports issued by the Swedish Data Inspection Board, the authority examined the treatment of personal data related to employer surveillance of employees. The reports indicated flaws occurring in regard to the Personal Data Act in several of the examined businesses.</p><p>In reference to this, a case study of a mid-sized call center was conducted to examine, from a technical point of view, the degree of compliance with the Personal Data Act regarding the company's surveillance of its employees. Based upon flaws found in the case study, a minor survey of similar companies was conducted in an attempt to measure the degree of generalizability of the identified flaws.</p><p>The flaws identified in the case study are mainly related to unsatisfactory routines for sorting out personal data during e-mail surveillance and lack of information given to employees about the conducting of checks related to this. A need to update the company's corporate IT policy and a need for a web surveillance solution were also noted.</p><p>The survey has indicated that similar flaws regarding screening routines and lack of information also might occur at other mid-sized call center companies.</p><p>The conclusion of the study is that several of the flaws the Swedish Data Inspection Board discovered five years ago still seem to be applicable among some mid-sized call center companies.</p>

O modelo de fila M/M/n+G pode ser usado para descrever o comportamento de uma Central de Atendimento. Nesse modelo as chegadas são Poisson com taxa lambda, o atendimento é exponencialmente distribuído com taxa mi, há n atendentes e os tempos de paciência dos clientes têm distribuição geral. A espera do usuário em fila não pode ultrapassar um tempo (paciência) que tem distribuição G e, se isto ocorrer, ele abandona o sistema. Mandelbaum e Zeltyn [2004] mostraram que existe uma relação linear entre o tempo médio de permanência na fila e a probabilidade de abandono nesses modelos quando a paciência é exponencialmente distribuída. No presente trabalho, estudamos essa relação no caso de distribuiçãao de paciência do tipo mista (com partes discreta e contínua), em que buscamos representar a reação dos usuários às mensagens gravadas reproduzidas periodicamente para aqueles que estão esperando atendimento. Utilizamos duas distribuições de paciência: Exponencial Mista e Uniforme Mista e percebemos que não há uma relação linear entre o tempo médio de espera na fila e a probabilidade de abandono. Observamos que para uma mesma taxa de chegada, o tempo médio de espera na fila é menor para a distribuição de paciência mista quando comparada com a Exponencial ou Uniforme de mesmos parâmetros. Analisamos o que ocorre com essa relação quando alteramos a distribuição do atendimento e percebemos que ela é mais afetada pela média e pelo coeficiente de variação do que pela particular distribuição escolhida para o tempo de serviço. / The M/M/n+G queueing model can be used to describe the behavior of a Call Center. This model has Poisson arrivals with rate lambda, service times are exponentially distributed with rate mi, n agents and the client´s patience time has general distribution. The waiting in line could not exceed a time (patience) which has distribution G, and if it occurs, the client leaves the system. In this models, Mandelbaum and Zeltyn [2004] showed that there is a linear relationship between average waiting time in queue and the probability of abandonment if the distribution of patience is Exponencial. In this work, we study this relationship in the case of patience with mixed distribution (which has discret and continuous parts). Through mixed distributions we try to represent the user´s reaction to recorded messages reproduced periodically when they are waiting for service. We have used Mixed Exponencial and Mixed Uniform distributions and, in both of them, there is not a linear relationship between average waiting time in queue and the probability of abandonment. We observe that for the same arrival rate, the average waiting time in line for mixed distribution is smaller than Exponencial or Uniform distributions with the same parameters. Also, we study the effect on waiting time and abandonment of different distributions of service and we observe that it is more affected by the coeficient of variation and average that by the particular distribution chosen for service.

Este trabalho estuda medidas de desempenho aproximadas em centrais de atendimento, apresentadas pelo modelo M/G/s/r+G. As aproximações são calculadas a partir do modelo M/M/s/r+M(n). Os resultados foram extendidos para o caso de mais de um tipo de cliente, apresentado pelo modelo M/Mi/s/r+Mi(n). Para dois casos particulares com 2 tipos de clientes, as aproximações citadas foram avaliadas numericamente e comparadas com os resultados de referência obtidos através de simulação. Os resultados aproximados comprovam que a aproximação é bastante satisfatória. / We study approximations for performance measures of call centers, represented by M/G/s/r+G queueing model. We use the measures computed in the M/M/s/r+M(n) queueing model. The results were extended for more then one type of customer, represented by M/Mi/s/r+Mi(n) queueing model. For two particular cases with two types of customers, the mentioned approaches were numerically evaluated and compared with the results of reference obtained by simulation. The approximate results show that the approach is quite satisfactory.

Submitted by Edisangela Bastos (edisangela@ufpa.br) on 2013-08-22T12:06:57Z No. of bitstreams: 2 Dissertacao_AvancosReconhecimentoFala.pdf: 830021 bytes, checksum: 66c2281f39ceb40eff9897cfe4299857 (MD5) license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) / Approved for entry into archive by Ana Rosa Silva(arosa@ufpa.br) on 2013-08-23T15:23:15Z (GMT) No. of bitstreams: 2 Dissertacao_AvancosReconhecimentoFala.pdf: 830021 bytes, checksum: 66c2281f39ceb40eff9897cfe4299857 (MD5) license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) / Made available in DSpace on 2013-08-23T15:23:15Z (GMT). No. of bitstreams: 2 Dissertacao_AvancosReconhecimentoFala.pdf: 830021 bytes, checksum: 66c2281f39ceb40eff9897cfe4299857 (MD5) license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) Previous issue date: 2013 / O reconhecimento automático de voz vem sendo cada vez mais útil e possível. Quando se trata de línguas como a Inglesa, encontram-se no mercado excelentes reconhecedores. Porem, a situação não e a mesma para o Português Brasileiro, onde os principais reconhecedores para ditado em sistemas desktop que já existiram foram descontinuados. A presente dissertação alinha-se com os objetivos do Laboratório de Processamento de Sinais da Universidade Federal do Pará, que é o desenvolvimento de um reconhecedor automático de voz para Português Brasileiro. Mais especificamente, as principais contribuições dessa dissertação são: o desenvolvimento de alguns recursos necessários para a construção de um reconhecedor, tais como: bases de áudio transcrito e API para desenvolvimento de aplicações; e o desenvolvimento de duas aplicações: uma para ditado em sistema desktop e outra para atendimento automático em um call center. O Coruja, sistema desenvolvido no LaPS para reconhecimento de voz em Português Brasileiro. Este alem de conter todos os recursos para fornecer reconhecimento de voz em Português Brasileiro possui uma API para desenvolvimento de aplicativos. O aplicativo desenvolvido para ditado e edição de textos em desktop e o SpeechOO, este possibilita o ditado para a ferramenta Writer do pacote LibreOffice, alem de permitir a edição e formatação de texto com comandos de voz. Outra contribuição deste trabalho e a utilização de reconhecimento automático de voz em call centers, o Coruja foi integrado ao software Asterisk e a principal aplicação desenvolvida foi uma unidade de resposta audível com reconhecimento de voz para o atendimento de um call center nacional que atende mais de 3 mil ligações diárias. / Automatic speech recognition has been increasingly more useful and feasible. When it comes to languages such as English, there are excellent speech recognizers available. However, the situation is not the same for Brazilian Portuguese, where the few recognizers for desktop dictation that existed, are no longer available. This dissertation is aligned with a goal of the Signal Processing Laboratory at the Federal University of Para, which is the development of a complete automatic speech recognizer for Brazilian Portuguese. More specifically, the main contributions of this dissertation are: the development of some resources needed to build a speech recognizer such as transcribed audio database and speech API; and the development of two applications: one for desktop dictation and another for automatic service in a call center. The system developed in-house for automatic speech recognition in Brazilian Portuguese is called Coruja, and besides all the resources that makes automatic speech recognition in Brazilian Portuguese available, the Coruja also contains an API for application development using speech recognition. The application for desktop dictation is called SpeechOO. The SpeechOO enables dictation and text editing and formatting by voice for the LibreOffice Writer. Other contribution of this work is the use of Coruja in call centers. Coruja was integrated with the Asterisk software, which is the main open source software for call centers. The main application developed for automated service in call center was an interactive voice response which is deployed nationally and receives more than 3 thousand daily calls.

O modelo de fila M/M/n+G pode ser usado para descrever o comportamento de uma Central de Atendimento. Nesse modelo as chegadas são Poisson com taxa lambda, o atendimento é exponencialmente distribuído com taxa mi, há n atendentes e os tempos de paciência dos clientes têm distribuição geral. A espera do usuário em fila não pode ultrapassar um tempo (paciência) que tem distribuição G e, se isto ocorrer, ele abandona o sistema. Mandelbaum e Zeltyn [2004] mostraram que existe uma relação linear entre o tempo médio de permanência na fila e a probabilidade de abandono nesses modelos quando a paciência é exponencialmente distribuída. No presente trabalho, estudamos essa relação no caso de distribuiçãao de paciência do tipo mista (com partes discreta e contínua), em que buscamos representar a reação dos usuários às mensagens gravadas reproduzidas periodicamente para aqueles que estão esperando atendimento. Utilizamos duas distribuições de paciência: Exponencial Mista e Uniforme Mista e percebemos que não há uma relação linear entre o tempo médio de espera na fila e a probabilidade de abandono. Observamos que para uma mesma taxa de chegada, o tempo médio de espera na fila é menor para a distribuição de paciência mista quando comparada com a Exponencial ou Uniforme de mesmos parâmetros. Analisamos o que ocorre com essa relação quando alteramos a distribuição do atendimento e percebemos que ela é mais afetada pela média e pelo coeficiente de variação do que pela particular distribuição escolhida para o tempo de serviço. / The M/M/n+G queueing model can be used to describe the behavior of a Call Center. This model has Poisson arrivals with rate lambda, service times are exponentially distributed with rate mi, n agents and the client´s patience time has general distribution. The waiting in line could not exceed a time (patience) which has distribution G, and if it occurs, the client leaves the system. In this models, Mandelbaum and Zeltyn [2004] showed that there is a linear relationship between average waiting time in queue and the probability of abandonment if the distribution of patience is Exponencial. In this work, we study this relationship in the case of patience with mixed distribution (which has discret and continuous parts). Through mixed distributions we try to represent the user´s reaction to recorded messages reproduced periodically when they are waiting for service. We have used Mixed Exponencial and Mixed Uniform distributions and, in both of them, there is not a linear relationship between average waiting time in queue and the probability of abandonment. We observe that for the same arrival rate, the average waiting time in line for mixed distribution is smaller than Exponencial or Uniform distributions with the same parameters. Also, we study the effect on waiting time and abandonment of different distributions of service and we observe that it is more affected by the coeficient of variation and average that by the particular distribution chosen for service.