- Home
- /
- ดาริชา สุธีวงศ์
รศ. ดร.ดาริชา สุธีวงศ์
- 6th Floor of Engineering 4 Bldg., Room 605
- +66-2218-6830
- daricha.s@chula.ac.th
Education
Ph.D. in Management Science and Engineering (Ph.D. minor in Computer Science)
Stanford University, United States, 2003
Master of Science in Engineering-Economic Systems and Operations Research
Stanford University, United States, 1997
Master of Engineering in Electrical Engineering and Computer Science
Massachusetts Institute of Technology, United States, 1996
Bachelor of Science in Computer Science and Engineering
Massachusetts Institute of Technology, United States, 1995
Expertise
Economics & Financial Engineering
Statistics & Data Analysis
Publications
2025
Passiri Bodhidatta, Daricha Sutivong
Understanding unnecessary stops and police use of force in NYPD Stop, Question, and Frisk with machine learning techniques Journal Article
In: Artificial Intelligence and Law, 2025, ISSN: 09248463, (Cited by: 0).
@article{Bodhidatta2025,
title = {Understanding unnecessary stops and police use of force in NYPD Stop, Question, and Frisk with machine learning techniques},
author = {Passiri Bodhidatta and Daricha Sutivong},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-86000624761&doi=10.1007%2fs10506-025-09444-y&partnerID=40&md5=cb643b7adb5d47c6b58bb0166c8c0e0e},
doi = {10.1007/s10506-025-09444-y},
issn = {09248463},
year = {2025},
date = {2025-01-01},
journal = {Artificial Intelligence and Law},
publisher = {Springer Nature},
abstract = {Even though the New York Police Department (NYPD) reform in 2013 led to a substantial reduction in the total number of stops, unnecessary stops and weapon use against innocent citizens remain critical issues. This study analyzes stop-and-frisk records during 2014 – 2019 using tree-based machine learning approaches along with logistic regression and Multi-Layer Perceptron (MLP) models, in order to discover patterns and insights. By developing predictive models for both suspect convictions and the level of force applied by police, this study provides a basis for a discussion whether weapon usage aligns with indicators of guilt or conviction. Findings show that XGBoost outperforms other machine learning techniques in predicting both conviction and the level of force used. Key factors associated with a suspect’s conviction include weapon possession, carrying suspicious objects, and trespassing. However, an excessive number of unnecessary stops appear to be associated with inaccurate assumptions about suspects’ weapon possession, which are also linked to police gunfire against innocent citizens. Refining suspicion criteria for Criminal Possession of Weapon and suspect actions could help reduce unnecessary stops and excessive force. © The Author(s), under exclusive licence to Springer Nature B.V. 2025.},
note = {Cited by: 0},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Even though the New York Police Department (NYPD) reform in 2013 led to a substantial reduction in the total number of stops, unnecessary stops and weapon use against innocent citizens remain critical issues. This study analyzes stop-and-frisk records during 2014 – 2019 using tree-based machine learning approaches along with logistic regression and Multi-Layer Perceptron (MLP) models, in order to discover patterns and insights. By developing predictive models for both suspect convictions and the level of force applied by police, this study provides a basis for a discussion whether weapon usage aligns with indicators of guilt or conviction. Findings show that XGBoost outperforms other machine learning techniques in predicting both conviction and the level of force used. Key factors associated with a suspect’s conviction include weapon possession, carrying suspicious objects, and trespassing. However, an excessive number of unnecessary stops appear to be associated with inaccurate assumptions about suspects’ weapon possession, which are also linked to police gunfire against innocent citizens. Refining suspicion criteria for Criminal Possession of Weapon and suspect actions could help reduce unnecessary stops and excessive force. © The Author(s), under exclusive licence to Springer Nature B.V. 2025.
2024
Korakot Kanjai, Kolratut Supbungerd, Ekkavich Chareonjirasak, Phattara Sripawatakul, Daricha Sutivong
Analyzing Effects of Time Series Data Characteristics on LSTM Performance Journal Article
In: Proceedings of the IEEE International Conference on Computer and Communications, ICCC, no. 2024, pp. 12 – 16, 2024, ISSN: 28377109, (Cited by: 0).
@article{Kanjai202412,
title = {Analyzing Effects of Time Series Data Characteristics on LSTM Performance},
author = {Korakot Kanjai and Kolratut Supbungerd and Ekkavich Chareonjirasak and Phattara Sripawatakul and Daricha Sutivong},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105006564553&doi=10.1109%2fICCC62609.2024.10941819&partnerID=40&md5=5eafc188aaa4233d172a4051079b2843},
doi = {10.1109/ICCC62609.2024.10941819},
issn = {28377109},
year = {2024},
date = {2024-01-01},
journal = {Proceedings of the IEEE International Conference on Computer and Communications, ICCC},
number = {2024},
pages = {12 – 16},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {The use of deep learning techniques for time series forecasting has increased in recent years. Long Short-Term Memory (LSTM) is a deep learning technique that has been shown to be effective for time series forecasting. This research proposes a methodology to study the relationship between LSTM performance and time series data characteristics, including trend, seasonality, skewness, kurtosis, chaotic behavior, and coefficient of variation. The methodology was applied to 48 stock price datasets. Three techniques, namely multiple linear regression, feature importance, and SHAP value, were used to compare LSTM performance across different data characteristics. The results show that coefficient of variation and chaotic behavior are the two most crucial characteristics affecting LSTM performance. © 2024 IEEE.},
note = {Cited by: 0},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The use of deep learning techniques for time series forecasting has increased in recent years. Long Short-Term Memory (LSTM) is a deep learning technique that has been shown to be effective for time series forecasting. This research proposes a methodology to study the relationship between LSTM performance and time series data characteristics, including trend, seasonality, skewness, kurtosis, chaotic behavior, and coefficient of variation. The methodology was applied to 48 stock price datasets. Three techniques, namely multiple linear regression, feature importance, and SHAP value, were used to compare LSTM performance across different data characteristics. The results show that coefficient of variation and chaotic behavior are the two most crucial characteristics affecting LSTM performance. © 2024 IEEE.
2023
Kanticha Wongnongtaey, Krissana Srisomboon, Daricha Sutivong
Institute of Electrical and Electronics Engineers Inc., 2023, ISBN: 979-835033386-2, (Cited by: 0).
@conference{Wongnongtaey2023130,
title = {Analyzing Suicide and Contributing Factors in Thailand Using Classification by Logistic Regression, Decision Tree and Support Vector Machine},
author = {Kanticha Wongnongtaey and Krissana Srisomboon and Daricha Sutivong},
editor = {Meen T.},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85166378651&doi=10.1109%2fICEIB57887.2023.10170197&partnerID=40&md5=79fd094b5cf5b6d789fd8376d567654c},
doi = {10.1109/ICEIB57887.2023.10170197},
isbn = {979-835033386-2},
year = {2023},
date = {2023-01-01},
journal = {2023 IEEE 3rd International Conference on Electronic Communications, Internet of Things and Big Data, ICEIB 2023},
pages = {130 – 135},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {The World Health Organization (WHO) reports that over 700,000 people die from suicide each year, with the number of attempts far exceeding that. Suicide is also the fourth leading cause of death among teenagers [1]. The suicide rate in Thailand follows a similar trend. This research aims to utilize machine learning techniques to identify cases that lead to suicide deaths and to study the contributing factors, both internal and external. Specifically, this paper proposes using classification methods to create a prediction model from people's demography, behaviors, and other factors in order to analyze features that influence the risk of suicide deaths. Classification techniques employed include Logistic Regression, Decision Tree, and Support Vector Machine. We have analyzed the data gathered by Thailand National Suicide Prevention Department from January 2018 to September 2021, which includes 49,417 suicidal cases, 30% of which resulted in fatalities. Two models are proposed: a general model to identify significant factors leading to suicide deaths and a practical model to predict suicide deaths given only available factors in practice. © 2023 IEEE.},
note = {Cited by: 0},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
The World Health Organization (WHO) reports that over 700,000 people die from suicide each year, with the number of attempts far exceeding that. Suicide is also the fourth leading cause of death among teenagers [1]. The suicide rate in Thailand follows a similar trend. This research aims to utilize machine learning techniques to identify cases that lead to suicide deaths and to study the contributing factors, both internal and external. Specifically, this paper proposes using classification methods to create a prediction model from people’s demography, behaviors, and other factors in order to analyze features that influence the risk of suicide deaths. Classification techniques employed include Logistic Regression, Decision Tree, and Support Vector Machine. We have analyzed the data gathered by Thailand National Suicide Prevention Department from January 2018 to September 2021, which includes 49,417 suicidal cases, 30% of which resulted in fatalities. Two models are proposed: a general model to identify significant factors leading to suicide deaths and a practical model to predict suicide deaths given only available factors in practice. © 2023 IEEE.
Sutiwat Simtharakao, Daricha Sutivong
Exploring Normalization Techniques in Neural Networks for Bitcoin Candlestick Price Prediction Conference
Institute of Electrical and Electronics Engineers Inc., 2023, ISBN: 978-166545645-6, (Cited by: 3).
@conference{Simtharakao2023483,
title = {Exploring Normalization Techniques in Neural Networks for Bitcoin Candlestick Price Prediction},
author = {Sutiwat Simtharakao and Daricha Sutivong},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85151995101&doi=10.1109%2fICAIIC57133.2023.10067086&partnerID=40&md5=cb23d63d58b547568a6fc720340cd519},
doi = {10.1109/ICAIIC57133.2023.10067086},
isbn = {978-166545645-6},
year = {2023},
date = {2023-01-01},
journal = {5th International Conference on Artificial Intelligence in Information and Communication, ICAIIC 2023},
pages = {483 – 488},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Bitcoin is a high-risk asset with a potentially high return. Predicting Bitcoin candlestick, i.e., open, high, low, and close (OHLC) prices, can help investors make trading decisions. This paper aims to explore various data normalization techniques in neural networks to enhance candlestick price prediction. In this study, the sliding window normalization techniques were compared with the whole set normalization techniques for forecasting the daily Bitcoin OHLC prices using two neural network algorithms: Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The investigated normalization techniques for both the whole set and the sliding window set included z-score normalization, min-max normalization, and relative change normalization. Historical OHLC prices over several days were used to predict the next day's OHLC prices. The results show that the sliding window normalization techniques outperformed the whole set normalization techniques in terms of RMSE and MAPE with the best technique being the GRU algorithm using the sliding window relative change normalization achieving MAPE of 2.25% and RMSE of 870.52. © 2023 IEEE.},
note = {Cited by: 3},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Bitcoin is a high-risk asset with a potentially high return. Predicting Bitcoin candlestick, i.e., open, high, low, and close (OHLC) prices, can help investors make trading decisions. This paper aims to explore various data normalization techniques in neural networks to enhance candlestick price prediction. In this study, the sliding window normalization techniques were compared with the whole set normalization techniques for forecasting the daily Bitcoin OHLC prices using two neural network algorithms: Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The investigated normalization techniques for both the whole set and the sliding window set included z-score normalization, min-max normalization, and relative change normalization. Historical OHLC prices over several days were used to predict the next day’s OHLC prices. The results show that the sliding window normalization techniques outperformed the whole set normalization techniques in terms of RMSE and MAPE with the best technique being the GRU algorithm using the sliding window relative change normalization achieving MAPE of 2.25% and RMSE of 870.52. © 2023 IEEE.
Nuttawan Sangsawai, Daricha Sutivong
Analyzing Impact of Economic Indicators on Vietnam Stock Market Using Machine Learning Techniques Conference
vol. 35, IOS Press BV, 2023, ISSN: 2352751X, (Cited by: 0; All Open Access, Gold Open Access).
@conference{Sangsawai2023279,
title = {Analyzing Impact of Economic Indicators on Vietnam Stock Market Using Machine Learning Techniques},
author = {Nuttawan Sangsawai and Daricha Sutivong},
editor = {Tang L.-C.},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85173465376&doi=10.3233%2fATDE230054&partnerID=40&md5=fa1718529d6236f9a2ec28cd477e415a},
doi = {10.3233/ATDE230054},
issn = {2352751X},
year = {2023},
date = {2023-01-01},
journal = {Advances in Transdisciplinary Engineering},
volume = {35},
pages = {279 – 288},
publisher = {IOS Press BV},
abstract = {Many people find interest in stock markets because of the potential financial gains. Understanding the fundamentals of each stock market is crucial as each has its own unique traits and driving factors. Traditionally, statistical methods are commonly used to find the relationship between various economic indicators and the stock markets. This study aims to utilize a different approach, namely machine learning techniques, a widely used tool for data analytics, to analyze the impact of economic indicators on the Vietnam stock index, which is a rising market during the past decade. The investigated machine learning algorithms include tree-based algorithms such as Decision Tree, Random Forest, and XGBoost. Monthly data, totaling 257 observations from August 2000 to December 2021, were used in this study. The results reveal that the XGBoost algorithm achieves the highest accuracy at 96.67% and the five most influential variables affecting the Vietnam stock market are S&P 500 index, consumer price index, exports, imports, and oil price, respectively, all with a positive relationship, while the relationships of the exchange rate, unemployment rate, and GDP with the Vietnam stock market are unclear. © 2023 The authors and IOS Press.},
note = {Cited by: 0; All Open Access, Gold Open Access},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Many people find interest in stock markets because of the potential financial gains. Understanding the fundamentals of each stock market is crucial as each has its own unique traits and driving factors. Traditionally, statistical methods are commonly used to find the relationship between various economic indicators and the stock markets. This study aims to utilize a different approach, namely machine learning techniques, a widely used tool for data analytics, to analyze the impact of economic indicators on the Vietnam stock index, which is a rising market during the past decade. The investigated machine learning algorithms include tree-based algorithms such as Decision Tree, Random Forest, and XGBoost. Monthly data, totaling 257 observations from August 2000 to December 2021, were used in this study. The results reveal that the XGBoost algorithm achieves the highest accuracy at 96.67% and the five most influential variables affecting the Vietnam stock market are S&P 500 index, consumer price index, exports, imports, and oil price, respectively, all with a positive relationship, while the relationships of the exchange rate, unemployment rate, and GDP with the Vietnam stock market are unclear. © 2023 The authors and IOS Press.
2020
Ponsuda Prutphongs, Daricha Sutivong
Decision Support System for Power Plant Improvement Investment Using Life-Cycle Cost Conference
Institute of Electrical and Electronics Engineers Inc., 2020, ISBN: 978-172818406-7, (Cited by: 0).
@conference{Prutphongs2020588,
title = {Decision Support System for Power Plant Improvement Investment Using Life-Cycle Cost},
author = {Ponsuda Prutphongs and Daricha Sutivong},
editor = {Wibowo F.W.},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85100018178&doi=10.1109%2fISRITI51436.2020.9315385&partnerID=40&md5=7c41d83838e1abb6a252980e80a2cbd4},
doi = {10.1109/ISRITI51436.2020.9315385},
isbn = {978-172818406-7},
year = {2020},
date = {2020-01-01},
journal = {2020 3rd International Seminar on Research of Information Technology and Intelligent Systems, ISRITI 2020},
pages = {588 – 592},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {This research designs and develops a Decision Support System (DSS) for evaluating a power plant improvement investment, given a generation plan, contract accounting and associated technical data. In practice, most owners often decide on an improvement investment by considering forward an immediate short period of revenue and expenses. This decision support system helps the owner take into account a more comprehensive period of cash flows in order to maximize the asset value and make an optimal decision. Specifically, the model calculation is based on the Life Cycle Cost Management (LCCM) under certain business rules. Our proposed evaluation model consists of five steps: 1) Consider the structure of revenue and expenses according to the Power Purchase Agreement (PPA). 2) Analyze accounting and technical data. 3) Estimate demand from the energy plan according to the system operator's yearly report. 4) Incorporate the data according to its business rule and the PPA constraints into the forecasting calculation. 5) Evaluate the investment using economic measures, such as Net Present Value (NPV) and Internal Rate of Return (IRR). © 2020 IEEE.},
note = {Cited by: 0},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
This research designs and develops a Decision Support System (DSS) for evaluating a power plant improvement investment, given a generation plan, contract accounting and associated technical data. In practice, most owners often decide on an improvement investment by considering forward an immediate short period of revenue and expenses. This decision support system helps the owner take into account a more comprehensive period of cash flows in order to maximize the asset value and make an optimal decision. Specifically, the model calculation is based on the Life Cycle Cost Management (LCCM) under certain business rules. Our proposed evaluation model consists of five steps: 1) Consider the structure of revenue and expenses according to the Power Purchase Agreement (PPA). 2) Analyze accounting and technical data. 3) Estimate demand from the energy plan according to the system operator’s yearly report. 4) Incorporate the data according to its business rule and the PPA constraints into the forecasting calculation. 5) Evaluate the investment using economic measures, such as Net Present Value (NPV) and Internal Rate of Return (IRR). © 2020 IEEE.
2018
Auth Pisutaporn, Burit Chonvirachkul, Daricha Sutivong
Relevant factors and classification of student alcohol consumption Conference
Institute of Electrical and Electronics Engineers Inc., 2018, ISBN: 978-153865696-9, (Cited by: 6).
@conference{Pisutaporn20181,
title = {Relevant factors and classification of student alcohol consumption},
author = {Auth Pisutaporn and Burit Chonvirachkul and Daricha Sutivong},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85049886823&doi=10.1109%2fICIRD.2018.8376297&partnerID=40&md5=b824d8e98341a3e07037b73ba8cc9141},
doi = {10.1109/ICIRD.2018.8376297},
isbn = {978-153865696-9},
year = {2018},
date = {2018-01-01},
journal = {2018 IEEE International Conference on Innovative Research and Development, ICIRD 2018},
pages = {1 – 6},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Educational data mining is the process of applying data mining tools and techniques to analyze data for educational purpose. This paper carries out educational data mining to study the student alcohol consumption through a public dataset which includes student attributes and their grades. The decision tree algorithm and the random forest algorithm are applied to perform classification and to analyze the variable importance. The regression model is then employed to illustrate the relationship between alcohol consumption level and the students' final grades. Our analysis provides knowledge on the relationship between student characteristics and alcohol consumption. The study also compares performance of the decision tree algorithm and the random forest algorithm. © 2018 IEEE.},
note = {Cited by: 6},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Educational data mining is the process of applying data mining tools and techniques to analyze data for educational purpose. This paper carries out educational data mining to study the student alcohol consumption through a public dataset which includes student attributes and their grades. The decision tree algorithm and the random forest algorithm are applied to perform classification and to analyze the variable importance. The regression model is then employed to illustrate the relationship between alcohol consumption level and the students’ final grades. Our analysis provides knowledge on the relationship between student characteristics and alcohol consumption. The study also compares performance of the decision tree algorithm and the random forest algorithm. © 2018 IEEE.
Suparerk Lekwijit, Daricha Sutivong
Optimizing the liquidity parameter of logarithmic market scoring rules prediction markets Journal Article
In: Journal of Modelling in Management, vol. 13, no. 3, pp. 736 – 754, 2018, ISSN: 17465664, (Cited by: 4).
@article{Lekwijit2018736,
title = {Optimizing the liquidity parameter of logarithmic market scoring rules prediction markets},
author = {Suparerk Lekwijit and Daricha Sutivong},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85053451144&doi=10.1108%2fJM2-06-2017-0066&partnerID=40&md5=8b00c0f1326db9d17a4ad0c266edccb0},
doi = {10.1108/JM2-06-2017-0066},
issn = {17465664},
year = {2018},
date = {2018-01-01},
journal = {Journal of Modelling in Management},
volume = {13},
number = {3},
pages = {736 – 754},
publisher = {Emerald Group Holdings Ltd.},
abstract = {Purpose: Prediction markets are techniques to aggregate dispersed public opinions via market mechanisms to predict uncertain future events’ outcome. Many experiments have shown that prediction markets outperform other traditional forecasting methods in terms of accuracy. Logarithmic market scoring rules (LMSR) is one of the most simple and widely used market mechanisms; however, market makers have to confront crucial design decisions including the setting of the parameter “b” or the “liquidity parameter” in the price functions. As the liquidity parameter has significant effects on the market performance, this paper aims to provide a comprehensive basis for the setting of the parameter. Design/methodology/approach: The analyses include the effects of the liquidity parameter on the forecast standard error and the amount of time for the market price to converge to the true value. These experiments use artificial prediction markets, the proposed simulation models that mimic real prediction markets. Findings: The simulation results indicate that prediction market’s forecast standard error decreases as the value of the liquidity parameter increases. Moreover, for any given number of traders in the market, there exists an optimal liquidity parameter value that yields appropriate price adaptability and leads to the fastest price convergence. Originality/value: Understanding these tradeoffs, the market makers can effectively determine the liquidity parameter value under various objectives on the standard error, the time to convergence and cost. © 2018, Emerald Publishing Limited.},
note = {Cited by: 4},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Purpose: Prediction markets are techniques to aggregate dispersed public opinions via market mechanisms to predict uncertain future events’ outcome. Many experiments have shown that prediction markets outperform other traditional forecasting methods in terms of accuracy. Logarithmic market scoring rules (LMSR) is one of the most simple and widely used market mechanisms; however, market makers have to confront crucial design decisions including the setting of the parameter “b” or the “liquidity parameter” in the price functions. As the liquidity parameter has significant effects on the market performance, this paper aims to provide a comprehensive basis for the setting of the parameter. Design/methodology/approach: The analyses include the effects of the liquidity parameter on the forecast standard error and the amount of time for the market price to converge to the true value. These experiments use artificial prediction markets, the proposed simulation models that mimic real prediction markets. Findings: The simulation results indicate that prediction market’s forecast standard error decreases as the value of the liquidity parameter increases. Moreover, for any given number of traders in the market, there exists an optimal liquidity parameter value that yields appropriate price adaptability and leads to the fastest price convergence. Originality/value: Understanding these tradeoffs, the market makers can effectively determine the liquidity parameter value under various objectives on the standard error, the time to convergence and cost. © 2018, Emerald Publishing Limited.
Pakom Walanaraya, Weerapat Puengpipattrakul, Daricha Sutivong
Movie Revenue Prediction Using Regression and Clustering Conference
Institute of Electrical and Electronics Engineers Inc., 2018, ISBN: 978-153866105-5, (Cited by: 4).
@conference{Walanaraya201863,
title = {Movie Revenue Prediction Using Regression and Clustering},
author = {Pakom Walanaraya and Weerapat Puengpipattrakul and Daricha Sutivong},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85054008064&doi=10.1109%2fICEI18.2018.8448610&partnerID=40&md5=29090e603d4640e185f0152141664dc4},
doi = {10.1109/ICEI18.2018.8448610},
isbn = {978-153866105-5},
year = {2018},
date = {2018-01-01},
journal = {2018 2nd International Conference on Engineering Innovation, ICEI 2018},
pages = {63 – 68},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Among many movies that have been released, some generate high profit while the others do not. This paper studies the relationship between movie factors and its revenue and build prediction models. Besides analysis on aggregate data, we also divide data into groups using different methods and compare accuracy across these techniques as well as explore whether clustering techniques could help improve accuracy. Specifically, two major steps were employed. Initially, linear regression, polynomial regression and support vector regression (SVR) were applied on the entire movie data to predict the movie revenue. Then, clustering techniques, such as by genre, using Expectation Maximization (EM) and using K-means were applied to divide data into groups before regression analyses are executed. To compare accuracy among different techniques, R-square and the root-mean-square error (RMSE) were used as a performance indicator. Our study shows that generally linear regression without clustering offers the model with the highest R-square, while linear regression with EM clustering yields the lowest RMSE. © 2018 IEEE.},
note = {Cited by: 4},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Among many movies that have been released, some generate high profit while the others do not. This paper studies the relationship between movie factors and its revenue and build prediction models. Besides analysis on aggregate data, we also divide data into groups using different methods and compare accuracy across these techniques as well as explore whether clustering techniques could help improve accuracy. Specifically, two major steps were employed. Initially, linear regression, polynomial regression and support vector regression (SVR) were applied on the entire movie data to predict the movie revenue. Then, clustering techniques, such as by genre, using Expectation Maximization (EM) and using K-means were applied to divide data into groups before regression analyses are executed. To compare accuracy among different techniques, R-square and the root-mean-square error (RMSE) were used as a performance indicator. Our study shows that generally linear regression without clustering offers the model with the highest R-square, while linear regression with EM clustering yields the lowest RMSE. © 2018 IEEE.
Pannate Jongpanichkultorn, Daricha Sutivong, Prabhas Chongstitvatana
Designing prediction markets to achieve convergence speed Journal Article
In: Engineering Journal, vol. 22, no. 4, pp. 177 – 190, 2018, ISSN: 01258281, (Cited by: 0; All Open Access, Gold Open Access, Green Open Access).
@article{Jongpanichkultorn2018177,
title = {Designing prediction markets to achieve convergence speed},
author = {Pannate Jongpanichkultorn and Daricha Sutivong and Prabhas Chongstitvatana},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85051290622&doi=10.4186%2fej.2018.22.4.177&partnerID=40&md5=db95bf5afbffd66eb3501e85aec0cc6c},
doi = {10.4186/ej.2018.22.4.177},
issn = {01258281},
year = {2018},
date = {2018-01-01},
journal = {Engineering Journal},
volume = {22},
number = {4},
pages = {177 – 190},
publisher = {Chulalongkorn University},
abstract = {The aim of this paper is twofold: to propose the model of artificial prediction markets that capture the characteristics of real prediction markets and to study the impact of key parameters on the performance of the proposed markets. In the experiments, the artificial markets are implemented and the market performance in terms of convergence speed is measured. Our experimental results show that the number of traders and the mean value of initial belief have no significant impact on the convergence speed. However, the trader’s memory size impacts negatively on the convergence because of its delay in adjusting to the true value. Finally, the external information transmission rate and the ratio of smart traders have positive impacts on the convergence of the prediction markets. The insights can assist a market maker in designing and constructing more efficient prediction markets. © 2018, Chulalongkorn University. All rights reserved.},
note = {Cited by: 0; All Open Access, Gold Open Access, Green Open Access},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The aim of this paper is twofold: to propose the model of artificial prediction markets that capture the characteristics of real prediction markets and to study the impact of key parameters on the performance of the proposed markets. In the experiments, the artificial markets are implemented and the market performance in terms of convergence speed is measured. Our experimental results show that the number of traders and the mean value of initial belief have no significant impact on the convergence speed. However, the trader’s memory size impacts negatively on the convergence because of its delay in adjusting to the true value. Finally, the external information transmission rate and the ratio of smart traders have positive impacts on the convergence of the prediction markets. The insights can assist a market maker in designing and constructing more efficient prediction markets. © 2018, Chulalongkorn University. All rights reserved.