## Mobimine: Monitoring the stock market from a PDA (2002)

### Cached

### Download Links

- [www.sigkdd.org]
- [www.sigkdd.org]
- [www.csee.umbc.edu]
- [www.cs.umbc.edu]
- [www.csse.monash.edu.au]
- [www.csee.umbc.edu]
- [www.cs.umbc.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | ACM SIGKDD Explorations |

Citations: | 28 - 8 self |

### BibTeX

@ARTICLE{Kargupta02mobimine:monitoring,

author = {Hillol Kargupta and Byung-hoon Park and Sweta Pittie and Lei Liu and Deepali Kushraj},

title = {Mobimine: Monitoring the stock market from a PDA},

journal = {ACM SIGKDD Explorations},

year = {2002},

volume = {3},

pages = {37--46}

}

### Years of Citing Articles

### OpenURL

### Abstract

### Citations

4978 |
C4.5: Programs for machine learning
- Quinlan
- 1993
(Show Context)
Citation Context ...n|the WatchList. 2.2.3 Data Mining The MobiMine makes use of a collection of online mining techniques including several statistical algorithms, clustering [15], Bayesian nets [10], and decision trees =-=[4; 25; 26]-=-. These techniques are primarily used in the Context component. The StockConnection module uses online statistical, Fourier spectrum-based decision tree and Bayesian learning techniques for detecting ... |

3945 |
Classification and regression trees
- Breiman, Friedman, et al.
- 1984
(Show Context)
Citation Context ...n|the WatchList. 2.2.3 Data Mining The MobiMine makes use of a collection of online mining techniques including several statistical algorithms, clustering [15], Bayesian nets [10], and decision trees =-=[4; 25; 26]-=-. These techniques are primarily used in the Context component. The StockConnection module uses online statistical, Fourier spectrum-based decision tree and Bayesian learning techniques for detecting ... |

3375 | Induction of decision trees
- Quinlan
- 1986
(Show Context)
Citation Context ...n|the WatchList. 2.2.3 Data Mining The MobiMine makes use of a collection of online mining techniques including several statistical algorithms, clustering [15], Bayesian nets [10], and decision trees =-=[4; 25; 26]-=-. These techniques are primarily used in the Context component. The StockConnection module uses online statistical, Fourier spectrum-based decision tree and Bayesian learning techniques for detecting ... |

2162 |
Algorithms for clustering data
- Jain, Dubes
- 1988
(Show Context)
Citation Context ...nt in order to create its personalized version|the WatchList. 2.2.3 Data Mining The MobiMine makes use of a collection of online mining techniques including several statistical algorithms, clustering =-=[15]-=-, Bayesian nets [10], and decision trees [4; 25; 26]. These techniques are primarily used in the Context component. The StockConnection module uses online statistical, Fourier spectrum-based decision ... |

1495 | Probability inequalities for sums of bounded random variables
- Hoeffding
- 1963
(Show Context)
Citation Context ... for information gain) at each decision node. Some of the recent techniques designed for large-scale applications include the Bootstrapping-based BOAT [11] and the Hoeffding (additive Chernoff bound) =-=[13]-=- tree-based VFDT [8] and the CVFDT [14]. An ensemble-based approach is proposed in [9]. It works using a Boosting-based approach to create ensemble of models. Different trees are generated from differ... |

1109 |
The Econometrics of Financial Markets
- Campbell, Lo, et al.
- 1997
(Show Context)
Citation Context ...ail system for enclosing the audio clips and reports. 2.1 MobiMine: What It is Not A large body of work exists that addresses dierent aspects of stock forecasting [1; 2; 7; 18; 20; 32] and selection [=-=6; 1-=-6] problem. The MobiMine is fundamentally dierent from the existing systems for stock forecasting and selection. First of all, it is dierent on the basis of philosophical point of view. In a tradition... |

597 | Bayesian network classifiers
- Friedman, Geiger, et al.
- 1997
(Show Context)
Citation Context ... its personalized version--the WatchList. 2.2.3 Data Mining The MobiMine makes use of a collection of online mining techniques including several statistical algorithms, clustering [15], Bayesian nets =-=[10]-=-, and decision trees [4; 25; 26]. These techniques are primarily used in the Context component. The StockConnection module uses online statistical, Fourier spectrum-based decision tree and Bayesian le... |

541 |
Portfolio Selection: Efficient Diversification of Investments
- Markowitz
- 1959
(Show Context)
Citation Context ...k and the desired level of return of investment are some of the key distinguishing features. The interestingness of a stock depends on these parameters. The MobiMine Client performs a Markowitz-style =-=[22]-=- optimization in order to personalize the list of interesting stocks for every user. Let ri be the return of investment for the i-th stock estimated over the period of observation and ffi be its norma... |

295 | Mining high-speed data streams
- Domingos, Hulten
- 2000
(Show Context)
Citation Context ...in) at each decision node. Some of the recent techniques designed for large-scale applications include the Bootstrapping-based BOAT [11] and the Hoeding (additive Cherno bound) [13] tree-based VFDT [8] and the CVFDT [14]. An ensemble-based approach is proposed in [9]. It works using a Boosting-based approach to create ensemble of models. Dierent trees are generated from dierent blocks of data obs... |

281 |
Constant depth circuits, Fourier transform, and learnability
- Linial, Mansour, et al.
- 1989
(Show Context)
Citation Context ...iscretization of a continuous space, we can compute its Fourier transformation. It turns out that the Fourier representation of a decision tree with bounded depth has some very interesting properties =-=[19; 21; 23]-=-. These observations are discussed in the following section. 3.3 Fourier Spectrum of a Bounded Depth Decision Tree For most practical applications decision trees have bounded depths. The Fourier spect... |

253 | Mining time-changing data streams
- Hulten, Spencer, et al.
- 2001
(Show Context)
Citation Context ...on node. Some of the recent techniques designed for large-scale applications include the Bootstrapping-based BOAT [11] and the Hoeding (additive Cherno bound) [13] tree-based VFDT [8] and the CVFDT [14]. An ensemble-based approach is proposed in [9]. It works using a Boosting-based approach to create ensemble of models. Dierent trees are generated from dierent blocks of data observed at dierent t... |

164 | Incremental induction of decision trees
- Utgoff
- 1989
(Show Context)
Citation Context ...mobile platforms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier efforts include ID4[27], ID5[29], ID5R=-=[30]-=- and ITI[31]. All these systems work using the ID3 style "information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally build a decision tre... |

116 | A streaming ensemble algorithm (sea) for large-scale classification
- Street, Kim
- 2001
(Show Context)
Citation Context ... the means to learn from large data sets and stream data [3; 5]. It is also based on adaptive re-sampling. However, it uses unweighted average to build the ensemble classier. Recently Street and Kim [=-=2-=-8] proposed Streaming Ensemble Algorithm (SEA) that learns an ensemble of decision trees for large-scale classication. SEA maintains asxed number of classiers. Once the ensemble becomes full, the k-th... |

112 |
Probability inequalities for sums of bounded random variables
- Hoeding
- 1963
(Show Context)
Citation Context ...re for information gain) at each decision node. Some of the recent techniques designed for large-scale applications include the Bootstrapping-based BOAT [11] and the Hoeding (additive Cherno bound) [1=-=-=-3] tree-based VFDT [8] and the CVFDT [14]. An ensemble-based approach is proposed in [9]. It works using a Boosting-based approach to create ensemble of models. Dierent trees are generated from dieren... |

102 | BOATâ€”optimistic decision tree construction
- Gehrke, Ganti, et al.
- 1999
(Show Context)
Citation Context ...t a time by keeping necessary statistics (measure for information gain) at each decision node. Some of the recent techniques designed for large-scale applications include the Bootstrapping-based BOAT =-=[1-=-1] and the Hoeding (additive Cherno bound) [13] tree-based VFDT [8] and the CVFDT [14]. An ensemble-based approach is proposed in [9]. It works using a Boosting-based approach to create ensemble of mo... |

81 |
Neural Network Time Series Forecasting of Financial Markets
- Azoff
- 1994
(Show Context)
Citation Context ...e interface can also invoke the e-mail system for enclosing the audio clips and reports. 2.1 MobiMine: What It is Not A large body of work exists that addresses different aspects of stock forecasting =-=[1; 2; 7; 18; 20; 32]-=- and selection [6; 16] problem. The MobiMine is fundamentally different from the existing systems for stock forecasting and selection. First of all, it is different on the basis of philosophical point... |

71 |
ID5: An incremental ID3
- Utgoff
- 1988
(Show Context)
Citation Context ...table for mobile platforms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier efforts include ID4[27], ID5=-=[29]-=-, ID5R[30] and ITI[31]. All these systems work using the ID3 style "information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally build a de... |

55 |
Beyond incremental processing: tracking concept drift
- Schlimmer, Granger
- 1986
(Show Context)
Citation Context ...hat is suitable for mobile platforms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier eorts include ID4[=-=27], ID5-=-[29], ID5R[30] and ITI[31]. All these systems work using the ID3 style \information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally b... |

52 | Time series segmentation for context recognition in mobile devices
- Himberg, Korpiaho, et al.
(Show Context)
Citation Context ... consuming data analysis tasks while on the move. So construction of the focus area seems to be an ideal mobile application for the data mining technology. Mining data streams in a mobile environment =-=[12; 17-=-] oers several unique challenges: 1. handling the continuoussow of incoming data, 2. eciently representing and communicating the data mining models over the wireless network with limited bandwidth, 3.... |

46 | an improved Algorithm for Incremental Induction of Decision Tress
- Utgoff
- 1994
(Show Context)
Citation Context ...orms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier efforts include ID4[27], ID5[29], ID5R[30] and ITI=-=[31]-=-. All these systems work using the ID3 style "information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally build a decision tree using one ... |

34 | Pasting bites together for prediction in large data sets and on-line
- BREIMAN
- 1997
(Show Context)
Citation Context ...vals. The ensemble classier for the stream is dened by a weighted average of the outputs of these trees. Breiman proposed an arcing method as the means to learn from large data sets and stream data [3=-=; 5-=-]. It is also based on adaptive re-sampling. However, it uses unweighted average to build the ensemble classier. Recently Street and Kim [28] proposed Streaming Ensemble Algorithm (SEA) that learns an... |

32 | Pasting small votes for classification in large databases and on-line
- Breiman
- 1999
(Show Context)
Citation Context ...ls. The ensemble classifier for the stream is defined by a weighted average of the outputs of these trees. Breiman proposed an arcing method as the means to learn from large data sets and stream data =-=[3; 5]-=-. It is also based on adaptive re-sampling. However, it uses unweighted average to build the ensemble classifier. Recently Street and Kim [28] proposed Streaming Ensemble Algorithm (SEA) that learns a... |

26 | The application of Adaboost for distributed, scalable and on-line learning
- Fan, Stolfo, et al.
- 1999
(Show Context)
Citation Context ...for large-scale applications include the Bootstrapping-based BOAT [11] and the Hoeding (additive Cherno bound) [13] tree-based VFDT [8] and the CVFDT [14]. An ensemble-based approach is proposed in [9]. It works using a Boosting-based approach to create ensemble of models. Dierent trees are generated from dierent blocks of data observed at dierent time intervals. The ensemble classier for the s... |

25 |
Bayesian network classi
- Friedman, Geiger, et al.
- 1997
(Show Context)
Citation Context ...e its personalized version|the WatchList. 2.2.3 Data Mining The MobiMine makes use of a collection of online mining techniques including several statistical algorithms, clustering [15], Bayesian nets =-=[10]-=-, and decision trees [4; 25; 26]. These techniques are primarily used in the Context component. The StockConnection module uses online statistical, Fourier spectrum-based decision tree and Bayesian le... |

13 |
Incremental induction of decision trees
- Utgo
- 1989
(Show Context)
Citation Context ... mobile platforms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier eorts include ID4[27], ID5[29], ID5R[=-=30] and -=-ITI[31]. All these systems work using the ID3 style \information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally build a decision tre... |

11 |
A Fourier Analysis-Based Approach to Learn Classifier from
- Park, R, et al.
- 2001
(Show Context)
Citation Context ...g partitionsj. If the magnitude of some w j is very small compared to other coecients then we may consider the j-th partition to be insignicant and neglect its contribution. As pointed out elsewhere [=-=17; 23; 2-=-4], the Fourier representation can be easily extended to the domains where features are non-Boolean. Consider a domain dened by ` possibly non-Boolean features where the i-th feature can take i disti... |

9 |
Financial Prediction Using Neural Networks
- Zirilli
- 1997
(Show Context)
Citation Context ...he interface can also invoke the e-mail system for enclosing the audio clips and reports. 2.1 MobiMine: What It is Not A large body of work exists that addresses dierent aspects of stock forecasting [=-=1; 2; 7; 18; 20; 3-=-2] and selection [6; 16] problem. The MobiMine is fundamentally dierent from the existing systems for stock forecasting and selection. First of all, it is dierent on the basis of philosophical point o... |

8 |
An intelligent forecasting system of stock price using neural networks
- Baba, Kozaki
- 1992
(Show Context)
Citation Context ...he interface can also invoke the e-mail system for enclosing the audio clips and reports. 2.1 MobiMine: What It is Not A large body of work exists that addresses dierent aspects of stock forecasting [=-=1; 2; 7; 18; 20; 3-=-2] and selection [6; 16] problem. The MobiMine is fundamentally dierent from the existing systems for stock forecasting and selection. First of all, it is dierent on the basis of philosophical point o... |

8 |
Portfolio Selection { Ecient Diversi of Investments
- Markowitz
- 1959
(Show Context)
Citation Context ...k and the desired level of return of investment are some of the key distinguishing features. The interestingness of a stock depends on these parameters. The MobiMine Client performs a Markowitz-style =-=[22-=-] optimization in order to personalize the list of interesting stocks for every user. Let r i be the return of investment for the i-th stock estimated over the period of observation and i be its norm... |

7 |
Learning decision rees using Fourier spectrum
- Kushilevitz, Mansour
- 1991
(Show Context)
Citation Context ...iscretization of a continuous space, we can compute its Fourier transformation. It turns out that the Fourier representation of a decision tree with bounded depth has some very interesting properties =-=[19; 21; 23]-=-. These observations are discussed in the following section. 3.3 Fourier Spectrum of a Bounded Depth Decision Tree For most practical applications decision trees have bounded depths. The Fourier spect... |

7 |
ID-5: An incremental ID-3
- Utgo
- 1988
(Show Context)
Citation Context ...itable for mobile platforms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier eorts include ID4[27], ID5[=-=29], ID5-=-R[30] and ITI[31]. All these systems work using the ID3 style \information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally build a de... |

7 |
An improved algorithm for incremental induction of decision trees
- Utgo
- 1994
(Show Context)
Citation Context ...forms connected over a wireless network. There exist several known techniques to construct incremental decision tree-based models. Some of the earlier eorts include ID4[27], ID5[29], ID5R[30] and ITI[=-=31]. All-=- these systems work using the ID3 style \information gain" measure to select attributes (or, equivalently, decision nodes). They are all designed to incrementally build a decision tree using one ... |

6 |
Mining time-critical data stream using the Fourier spectrum of decision trees
- Kargupta, Park
- 2001
(Show Context)
Citation Context ... consuming data analysis tasks while on the move. So construction of the focus area seems to be an ideal mobile application for the data mining technology. Mining data streams in a mobile environment =-=[12; 17-=-] oers several unique challenges: 1. handling the continuoussow of incoming data, 2. eciently representing and communicating the data mining models over the wireless network with limited bandwidth, 3.... |

6 |
Knowledge Discovery from Heterogeneous Data Streams Using Fourier Spectrum of Decision Trees
- Park
- 2001
(Show Context)
Citation Context ...ourier basis function that depends on a certain subset of features dening the domain of the data set to be mined. The entire process of converting decision trees to a Fourier Spectrum is detailed in [=-=23-=-]. The next section oers a short introduction to the multi-variate Fourier basis representation. 3.2 A Brief Review of the Fourier Basis Consider the function space over the set of all `-bit Boolean f... |

5 |
A neural network approach for forecasting and analyzing the price-volume relationship in the taiwan stock market
- Cheng
- 1994
(Show Context)
Citation Context ...he interface can also invoke the e-mail system for enclosing the audio clips and reports. 2.1 MobiMine: What It is Not A large body of work exists that addresses dierent aspects of stock forecasting [=-=1; 2; 7; 18; 20; 3-=-2] and selection [6; 16] problem. The MobiMine is fundamentally dierent from the existing systems for stock forecasting and selection. First of all, it is dierent on the basis of philosophical point o... |

5 |
Intelligent stock trading decision support system using dual adaptive-structure neural networks
- Jang, Lsi, et al.
- 1993
(Show Context)
Citation Context ...ail system for enclosing the audio clips and reports. 2.1 MobiMine: What It is Not A large body of work exists that addresses dierent aspects of stock forecasting [1; 2; 7; 18; 20; 32] and selection [=-=6; 1-=-6] problem. The MobiMine is fundamentally dierent from the existing systems for stock forecasting and selection. First of all, it is dierent on the basis of philosophical point of view. In a tradition... |

4 |
Pasting small votes for classi in large databases and on-line
- Breiman
- 1999
(Show Context)
Citation Context ...vals. The ensemble classier for the stream is dened by a weighted average of the outputs of these trees. Breiman proposed an arcing method as the means to learn from large data sets and stream data [3=-=; 5-=-]. It is also based on adaptive re-sampling. However, it uses unweighted average to build the ensemble classier. Recently Street and Kim [28] proposed Streaming Ensemble Algorithm (SEA) that learns an... |

4 |
Intelligent stock market forecasting system through artificial neural networks and fuzzy delphi
- Lee
- 1996
(Show Context)
Citation Context |

3 |
Neural Network Time Series Forecasting of Financial Markets
- Azo
- 1994
(Show Context)
Citation Context |