Luke Luke Gebauer Gebauer College: C ll C ll College: Mentor: Mentor: Allan Hancock Allan Hancock Kathy Macropol Kathy Macropol Major: Major: Major: Major: Advisor: Advisor: Electrical Engineer Electrical Engineer Dr. Dr. Ambuj Ambuj Singh Singh Computer Computer Science Science F F Funded Funded by d d d d b by the the US Army US Army
Sales Sales 1st Qtr 2nd Qtr Q Y Value 1 Y ‐ Value 1 3rd Qtr 3.5 4th Qtr 3 2 5 2.5 2 6 1.5 5 1 1 4 Series 1 0.5 3 Series 2 0 Series 3 2 0 1 2 3 1 0
Wh What makes it time Wh What makes it time ‐ evolving k k i i i i evolving l i l i Los Traffic Data Traffic Data Reno Angeles Time: 11:00am 5:00pm 2:00am p Los Traffic: 100 Vehicles 50 Vehicles 900 Vehicles Santa San Nodes Nodes Francisco Angeles Angeles Barbara Barbara HWY 101 Santa Edges Edges Barbara Barbara Stockton Bakersfield Bakersfield Data Mining Tasks Data Mining Tasks
Research Objectives Research Objectives Research Objectives Research Objectives Importance Importance Provide additional Provide additional time evolving graphs for future data Construct a time mining algorithms evolving graph of countries and their Predict the future international trade international trade P Perform clustering f l t i socio ‐ economic transactions algorithms on statuses and constructed graph international relations of countries
Clusters Edges Nodes Nodes Clustering Algorithm Preprocessing Graph
Large scale image of Large scale image of Large scale image of Large scale image of time ‐ evolving country time ‐ evolving country graph graph Enlarged image of time ‐ evolving of time evolving country graph
849 849 Clustered Clustered Clustered Clustered 2693 2693 2693 2693 3994 3994 2734 2734 time time ‐ evolving evolving 7393 7393 1324 1324 country graph country graph country graph country graph 5921 5921 5921 5921 3929 3929 3929 3929 3294 3294 1193 1193 1246 1246 795 795 2734 2734 1293 1293 5394 5394 5394 5394 4203 4203 273 273 Enlarged image of Enlarged image of clustered graph clustered graph
Data is Data is Data is Data is Pearson Correlation Coefficient ‐ 0.2 significant! significant! (between the average physical distance of (significance is anything that has an the countries in the clusters, and the absolute value greater than ~0 07) absolute value greater than 0.07) average cluster score) l ) Conclusion Conclusion Since the cluster score increases as the average distance between countries decreases, greater amounts of trading takes place between countries that are in p close proximity to one another
Additional hypotheses to verify through data mining Additional hypotheses to verify through data mining ddi i ddi i l h l h h h if if h h h d h d i i i i Based upon their current trade Based upon their current trade h Do smaller countries tend Do smaller countries tend values, can a model be built to values, can a model be built to to trade in close to trade in close predict how clusters of countries will predict how clusters of countries will proximity? proximity? form in the future? form in the future? What other attributes of a country What other attributes of a country Do countries tend to Do countries tend to determine its trading values (i.e. determine its trading values (i.e. stay within their own stay within their own population, geographical placement, population, geographical placement, clusters? clusters? etc…)? etc…)?
Edges Edges Nodes Nodes Data objects that we Data objects that we Some common factor Some common factor want to know more between the data objects information about that changes value over time Predictive Tasks Predictive Tasks Descriptive Tasks Descriptive Tasks Models that can predict Deriving patterns that future events within future events within summarize the underlying summarize the underlying the graph based upon relationships between the its attributes data objects
¼ 1/3 ½ 1/3 ¼ 1/3 ½ 1/3 Build a probability Build a probability Initial probability Initial probability ¼ 1/3 0 1/3 ¼ 1/3 0 1/3 matrix and perform matrix and perform matrices matrices ¼ 0 ½ 0 ¼ 0 ½ 0 random walks using random walks using ¼ 1/3 0 1/3 ¼ 1/3 0 1/3 “Markov Chains” until “Markov Chains” until the converged matrix is the converged matrix is g “ “Markov Chains” Process “ “Markov Chains” Process k k h h ” ” obtained. obtained. Cluster structure Cluster structure 1 2 Will eventually equal Will eventually equal .50 .50 ‐‐ .50 the converged matrix the converged matrix ‐‐ ‐‐ ‐‐ ‐‐ where possible cluster where possible cluster p ‐‐ ‐‐ 1 0 ‐‐ 1.0 structures may be seen structures may be seen .50 .50 ‐‐ .50 3 4
Main Purpose Main Purpose Main Purpose Main Purpose Genes Genes Predict an unknown Predict an unknown Predict an unknown Predict an unknown gene’s function by gene’s function by gene’s function by gene’s function by looking at what looking at what looking at what looking at what Shared Gene Shared Gene other genes were other genes were other genes were other genes were other genes were other genes were other genes were other genes were Functions Functions Functions Functions clustered together clustered together clustered together clustered together with it based upon with it based upon with it based upon with it based upon k k known shared known shared k k known shared known shared h h h h d d d d functions functions functions functions
Further Data Mining Further Data Mining Further Data Mining Further Data Mining Granger Causality Granger Causality Granger Causality Granger Causality Analysis Analysis Analysis Analysis y Coefficient Coefficient Coefficient Coefficient Build predictive models for Build predictive models for Build predictive models for Build predictive models for Statistical method which Statistical method which Statistical method which Statistical method which graph based upon its attributes graph based upon its attributes graph based upon its attributes graph based upon its attributes graph based upon its attributes graph based upon its attributes graph based upon its attributes graph based upon its attributes measures how likely it is measures how likely it is measures how likely it is measures how likely it is h h h h lik l it i lik l it i lik l it i lik l it i that one property causes that one property causes that one property causes that one property causes Test to determine if there are Test to determine if there are Test to determine if there are Test to determine if there are another, across time another, across time another, across time another, across time any connections between any connections between any connections between any connections between properties of the countries properties of the countries properties of the countries properties of the countries Do changes in population Do changes in population Do changes in population Do changes in population within the clusters and how within the clusters and how within the clusters and how within the clusters and how drive changes in trade or drive changes in trade or drive changes in trade or drive changes in trade or drive changes in trade or drive changes in trade or drive changes in trade or drive changes in trade or those clusters form those clusters form those clusters form those clusters form cluster size? cluster size? cluster size? cluster size?
Recommend
More recommend
