DSpace Collection:
http://dspace.cityu.edu.hk:80/handle/2031/769
2016-10-26T23:39:25ZModeling methods for residential energy consumption and user behaviors of online-targeted display advertising
http://dspace.cityu.edu.hk:80/handle/2031/8151
Title: Modeling methods for residential energy consumption and user behaviors of online-targeted display advertising
Authors: Guan, Jingjing (關菁菁)
Abstract: This dissertation presents modeling methods for two research problems:
residential energy consumption (REC) and user click-through behaviors towards
online-targeted display advertising (OTDA).
Understanding REC via modeling household-level survey data is
important for governments, energy corporations, and home appliance
manufacturers. National residential energy consumption surveys (RECS) collect
household-level data with stratified random sampling schemes. RECS data,
consequently, have a natural and explicit multilevel structure, reflected by
geographical clustering of households. To handle this multilevel structure, we
introduce a multilevel regression model. This approach divides overall REC
variations into two sources: area and household-level variations; and respectively
explains them with environmental and household-level variables. With a 53%
explained variance proportion (EVP) (82% of area variations and 47% of
household-level variations); the proposed multilevel regression model
outperforms previous models, e.g. traditional linear regression models.
Furthermore, the multilevel regression model largely reduces the impact of area
variations on the estimated effects of influencing factors of REC.
We introduce regularized linear models (RLMs) with the elastic net
penalty to model REC. The purpose is to identify significant factors among all
utilizable variables from RECS micro-datasets. This approach imposes no
antecedent theory on variable selection. The elastic net penalty is a compromise of
the ridge-regression and the LASSO penalty. It helps to handle more than 500
RECS variables of complicated multicollinearity in one model. With the U.S.
2009 RECS dataset, we develop a series of RLMs with the elastic net penalty. All
constructed RLMs simultaneously assign non-zero effects to 98 selected variables.
The best-fit RLM, which explains 65% of the total variability, outperforms most
previous models in the literature.
OTDA, as a new mode of online display advertising, has developed
rapidly due to its capability to target potential customers. This dissertation
addresses the issue from the perspective of OTDA publishers. As many
management problems inherently involve optimization and statistical modeling, we develop a two-step forecasting method to forecast user click-through behaviors
towards OTDA, so as to control uncertainties in formulating allocation plans for
OTDA resources.
We introduce a finite mixture regression model, i.e., an arbitrary-pointsinflated
(API) Poisson regression model as a foreshadowing. With an offset in the
Poisson component, this model can handle count data with an arbitrary number of
inflated points and link clicks with page views. We develop algorithms for
parameter estimation, adaptively choosing the best-fit API Poisson regression
model according to the Bayesian information criterion (BIC), and obtaining the
corresponding Hessian matrix.
The two-step forecasting method involves a modeling and a predicting
step. It can forecast user clicks towards matches of advertising requests and
candidate allocation plans, based on data observed in current period. The
modeling step segments data in current period into sub-samples with an adequate
number of sample sizes, and constructs sub-models using an adaptive API Poisson
regression algorithm. The predicting step provides two predicting schemes, and
selects the scheme with higher per campaign prediction accuracy as the final
scheme, to forecast user clicks in next period. Moreover, the proposed method is
of fast computing speed and robust parameter estimation.
We adapt this two-step forecasting method to forecast user clicks towards
OTDA for a social network site. The empirical results show that our approaches
have higher accuracy than other previous methods, including logistic regression,
truncated logistic regression, etc. The ensemble-predicting scheme achieves
higher accuracy in forecasting non-zero clicks, compared to the campaign-tocampaign
predicting scheme. The model involving page views possesses the
smallest prediction error among all alternative models being considered. Finally,
we present a brief discussion on forecasting page views and suggest a further
extension of the API Poisson regression to model count data other than Poisson
distribution.
Notes: CityU Call Number: HB849.49 .G83 2014; xviii, 197 p. : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2014.; Includes bibliographical references (p. 177-195)2014-01-01T00:00:00ZVolatility surface, term structure and meta-learning-based price forecasting for option strategies design
http://dspace.cityu.edu.hk:80/handle/2031/8084
Title: Volatility surface, term structure and meta-learning-based price forecasting for option strategies design
Authors: Zhou, Shifei (周仕飛)
Abstract: The forecasting of underlying asset price is important for investors to make financial
decisions. A successful prediction can save investors from risk of losing money. This
thesis focuses on the forecasting of underlying asset price and develops an
option-based trading system. A literature review is conducted on volatility and its
related topics. These topics include volatility forecasting, implied volatility smile,
implied volatility term structure, implied volatility surface, local implied volatility and
stochastic volatility. The major forecasting models and methodologies of volatility
prediction are introduced and classified. This classification also gives a direct
blueprint for the composition of this thesis. Based on the investigation, this thesis
proposes three research topics and makes contributions as follows.
First, a model-free term structure-based stochastic model with adaptive
correlation is proposed for price forecasting. Based on observations, the constant
assumption of correlation of stochastic volatility model is found to be unsuitable for
analyzing Hong Kong options market. The least squares method is used to evaluate
this correlation. Besides, the term structure implied volatility is obtained by
integrating option price and strike price from current time to expiry date. This
model-free term structure is used as the long-run mean level of stochastic model to
make use of information contained in term structure. Empirical test shows our model
outperforms CEV model and Regression model in terms of one-day-ahead prediction
performance and 78-day distribution of underlying asset price.
Second, a novel local volatility model with mean-reversion process is proposed.
This mean-reversion term is functioned as long run mean level of local volatility
surface. The larger local volatility departs from its mean level, the greater rate local
volatility will be reverted with. Then, a B-spline with moving average knot control
scheme is applied to interpolate local volatility matrix. The bi-cubic B-spline is used to recover local volatility surface from this local volatility matrix. Finally, Monte
Carlo simulation is adopted to predict underlying asset price. Empirical tests show our
mean-reversion local volatility model has a good prediction performance than
traditional local volatility models.
Third, an improved EMD meta-learning rate-based model for gold price
forecasting is proposed. Firstly, we adopt the EMD method to divide the time series
data into different subsets. Secondly, a back-propagation neural network model
(BPNN) is used to function as the prediction model in our system. We update the
online learning rate of BPNN instantly as well as the weight matrix. Finally,
forecasting results from different BPNNs are summed as a final price forecasting
result. The experiment results show that our system has a good forecasting
performance.
Based on the above three theoretical innovation to current financial models, the
forecasting results of three different models are integrated by an average method as a
final forecasting price value. This value is used to decide the movement trend of
underlying asset price. According to the trend, six different movement patterns are
classified. The corresponding option trading strategies are also designed. Then, the
optimal option trading strategy is selected by three criteria. There are Expected Return,
Value at Risk, and Conditional Value at Risk.
To sum up, this thesis proposes three different models to forecast price and
designs option trading strategies based on three criteria. The future works contain two
aspects. First, the system will be improved for high frequency trading. The
improvement includes calculation optimization and model optimization. Second, the
system will be applied to other options and futures markets.
Notes: CityU Call Number: HG6024.A3 Z4948 2013; x, 156 p. : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2013.; Includes bibliographical references (p. 149-156)2013-01-01T00:00:00ZCombinatorial optimization techniques for manpower scheduling problems
http://dspace.cityu.edu.hk:80/handle/2031/8083
Title: Combinatorial optimization techniques for manpower scheduling problems
Authors: Zhang, Zizhen (張子臻)
Abstract: This thesis studies a class of manpower scheduling problems in the field of
operations research. In daily operations of a company, we may need to dispatch
a number of employees to carry out a set of requests or tasks. Making a good
schedule for each employee is one of the most crucial decisions for the manager
and decision maker. Generally speaking, a schedule is a subset of tasks assigned
to employees over a specific planning horizon. The high-quality schedule can
help improve the efficiency of manpower and reduce the operational cost, thereby
strengthening the competitiveness of the company.
The manpower scheduling problems we investigate in this thesis aim at planning
a schedule that can satisfy the following main requirements. First, a schedule
must be practical. It needs to take various kinds of rules or regulations into account.
Second, it should provide sufficient
exibility for the employees to perform
their assigned tasks. Third, the company can reduce their operational cost without
the provided service level being negatively affected.
Apart from the manpower scheduling problems, there are many other scheduling
problems in literature. For example, machine scheduling problems and jobshop
scheduling problems are two typical types of scheduling problems that have attracted much academic attention in the last several decades. Both of them
share some similarities with the manpower scheduling problems. However, the
manpower scheduling problems put more emphasis on the nature and behavior
of human beings. In particular, we summarize some characteristics of the manpower
scheduling problems as follows: (1) working periods and working hours;
(2) skills and qualifications; (3) travel and service time; (4) collaborations and
synchronization; (5) knowledge and learnability; and (6) other characteristics.
In this thesis, we first introduce the background of our study and then present
some related literature of the manpower scheduling problems in the area of
discrete combinatorial optimization. Subsequently, we elaborate three practical
manpower scheduling problems, namely, the Manpower Scheduling Problem
with Crew Collaboration Constraints (MSPCC), the Manpower Scheduling Problem with Regular Working Hour Constraints (MSPRWH), and the Manpower
Scheduling Problem with Crew Holding Cost (MSPCHC). All these problems are
stemmed from real-world applications, which re
ect some or all above-mentioned
characteristics. Finally, we conclude the thesis with closing remarks.
The MSPCC is a practical scheduling and routing problem that tries to synchronize
worker schedules to complete all tasks. We first provide an integer
programming model for the problem and discuss its properties. Next, we show
that the tree data structure can be used to represent the MSPCC solutions, and
its optimal solution can be obtained from one of trees by solving a minimum
cost
ow model for each worker type. Based on the above findings, we develop
for the problem a novel tabu search algorithm employing tree-based search operators.
Finally, we evaluate the effectiveness of the tabu search algorithm by
computational experiments on two sets of instances.
The MSPRWH examines an inspector scheduling problem with time windows
whereby labour regulations affect planning horizons, and therefore, profitability. The goal is to determine a schedule, that is, a set of routes for inspectors that
maximizes profitability from visited locations, based on the conditions that inspectors
can only travel during stipulated working hours within each period in a
given planning horizon and that the inspectors are only required to return to the
depot at the end of the last period. We propose an effective tabu search algorithm
with an ejection pool representation to solve the MSPRWH. We evaluate the
effectiveness of our tabu search algorithm with extensive experiments based on
the team orienteering problem with time windows instances and a set of modified
solomon's benchmark instances. The results indicate that our approach generates
high-quality solutions.
The MSPCHC is a simplified version of the real-world film shooting problem,
which aims to determine a shooting sequence so as to minimize the total cost of the
actors involved. We first formulate the problem as an integer linear programming
model and then devise a branch-and-bound algorithm to solve it. Subsequently,
we enhance the branch-and-bound algorithm by several accelerating techniques,
including preprocessing, dominance rules and caching search states. Extensive
experiments over two sets of benchmark instances suggest that our branch-and-bound
algorithm is superior to the currently best exact algorithm for the problem.
Finally, the impacts of different parameter settings are also disclosed by some
additional experiments.
Notes: CityU Call Number: HF5549.5.M3 Z45 2014; xii, 137 p. : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2014.; Includes bibliographical references (p. 122-137)2014-01-01T00:00:00ZEssays on supply chain finance
http://dspace.cityu.edu.hk:80/handle/2031/8082
Title: Essays on supply chain finance
Authors: Wu, Kekun (吳克坤)
Abstract: This thesis mainly studies two financing problems in the context of supply
chain management. The first problem is motivated by a major agricultural
firm's practice in China. We consider an agribusiness model called
firm+farmers". We aim to build a theoretical foundation for this kind
of business models that have attracted growing attention from the industry,
government, and the farmers in developing economies such as China.
The involving firm offers a revenue-sharing contract to a large number of
farmers, many of whom are tiny in nature. The farmers, based on their
expected returns and risks, choose to accept or reject the contracts. Our
model is distinguished from the previous literature on contract farming and
agricultural operations management in the following way. The contract on
the one hand retains a revenue sharing mechanism between the firm and
the farmers, and on the other hand guarantees a minimum payment to the
farmers. Thus, neither the firm nor the farmers bear full production and
market risks but the firm takes greater risks. We analyze the optimal supply
chain decisions for the firm and the farmers under such contracts. We
show how the business model provides financing benefits to both the firm
and the farmers that may otherwise not be accessible for them. We explain
how the pooling effect may influence the supply chain decisions and
performance. Our results address for the first time that how such hybrid
contracts may shape the agricultural supply chain and help drive sustainable development, and uncover the underling theoretical mechanism and
merits behind the apparent success of these agribusinesses.
In the second problem, we study a capacity procurement problem faced
by a shipping company whose financial resource is costly. We propose a
capacity-cost sharing scheme, which, commonly observed in the industry,
is to allow the company to share the capacity and the procurement cost
of a ship. A first-price auction is implemented to assist the company in
determining the percentage of capacity and cost to share. Our analysis
shows that first-price auction yields more revenue compared to a secondprice
auction, which is distinct from the previous work on auction. The
essay contributes to the theory of auction in that the the bidder's signal
space has more than one dimension. We use isobid curves to reduce the
dimensionality of the signal space. An efficient heuristic is designed to
compute the bidder's equilibrium bids numerically.
Notes: CityU Call Number: HD38.5 .W77 2013; ix, 87 p. : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2013.; Includes bibliographical references (p. 83-87)2013-01-01T00:00:00Z