MIS Reading Notes 1-14-16
MIS Reading Notes 1-14-16 MIS 0855-004
Popular in Data Science
Popular in Business, management
This 3 page Class Notes was uploaded by Rebecca Jackson on Saturday February 6, 2016. The Class Notes belongs to MIS 0855-004 at Temple University taught by Laurel Miller in Spring 2016. Since its upload, it has received 8 views. For similar materials see Data Science in Business, management at Temple University.
Reviews for MIS Reading Notes 1-14-16
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 02/06/16
Data Science and Prediction Predictive Modeling Science: a systematic enterprise that builds and organizes knowledge in the form of testable explanations and predictions (data science is not statistics) Huge amounts of unstructured raw data, tagged for computers to interpret and make decisions of Looking for predictive patterns in massive amounts of data, rather than looking for data to satisfy a pattern Predictively actionable Simple theories are useless but common b/c they are likely to have predictive accuracy; explanations are weak, bold but falsifiable better. Implications o Storage almost free: everything is stored in case it is useful (videos, social media messages) o 1980s/90s became popular, software developed to use data for prediction o Machine learning picks up subtle structure in data, but also “noise”; no assumptions made on relationship o Skips the assumptive theories; gathers patterns without guessing an explanation o Helpful when there are many possibilities (is it x, is it y?) o Aggregated: gathered. To find patterns aggregate data even if it is exaggerating similarities o From data, learning algorithm finds pattern o But would that pattern hold up predictively? o Oddly specific patterns helpful (health industry warnings) o In large varied data, difficult to imagine patterns. Simpler is better if its more predictive o Computers can test for predictive accuracy across large amounts of data o Causation can be found but background of data must be known and all variables involved should be included Skills o Machine learning, test processing/mining, XML tagging (unstructured data) o 3 Basic Skills Statistics, econometrics, despite their basis on fitting data to linear models Computer science, systems skills, computer languages, despite limited to moderate data Correlation and Causation, causal desirable but tricky to get o Ability to formulate problems in a way that leads to efective solutions; isomorphism, identical underlying structure o “Unbalanced target classes” – dependent variable is interesting only sometimes, difficult to predict o “computational thinking” – problem formulation skills o Shift to data-driven decision making o Computer>Human: cost, scalability and accuracy Knowledge Discovery o No longer need to settle for models or theories; let machine learning get it all right Can science & models exist without theories? “concept drift: - models get worse over time, predictive data is updated Biases unless intended for finding causality Good for where we can be inaccurate; NOT PHYSICS. Earth/social science Social very inaccurate, incomplete, come up with causality & gather data but same data can = two diferent theories easily Even if cause unknown, predictive accuracy can save lives (health), find cause later Model errors: misspecification (linear model on nonlinear data), sample size, heterogeneous data Big data = more data to test models w/reliable error bounds, good proxy for pop No “clean data”, no controls or ability to know anything but passive data Space to test some theories w/controlled experiments w/internet Predictive models = how everyone will vote and how to change them Social science theories ^ due to big interaction on net Conclusion o Big data can look at situations that wouldn’t occur in controlled experiments o Machine asks questions humans won’t think of, large & varied wealth of data o Computer science, statistics, causal modeling… o Machine learning & predictive modeling; machine learning + human knowledge o Orgs: Test on internet & change based on data Three Science Words Hypothesis : testable prediction from an idea Theory : supported by evidence Scientific Law : generally true Model : replace them all, remaking of something else o Physical Model o Mathematical Model o Conceptual Model
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'