syllabus.
- 1. System Quality Fall 2002 Professor Arthur Goldberg Schedule (version 9.11.02) Week Date Topic Description Readings Assignments 1 9/4 System quality Problems caused by lousy computer systems. What is system quality? (ISO 9126 Buggy software still takes a toll, introduction; handout). Software and data quality. Business software firms sued over Software Syllabus overview. Readings. Assignments. implementation, Development Best Software quality: Software development approaches. Best practices. Peer reviews Gridlock as 800 traffic lights seize, and software testing. Fishermen rescued after dam malfunction Practices, Capers Data quality: data ownership, data analysis tools and techniques, metadata, data http://catless.ncl.ac.uk/Risks/22.14.html#subj2 Jones, Part I quality rules and data cleansing. , Software problem kills soldiers in training incident http://catless.ncl.ac.uk/Risks/22.13.html#subj2 2 9/11 Software Software development best practices, Software development quality, benchmark Capers Jones, Sizing Up Software, Scientific Development Best measures of productivity and defect rates, Benchmark studies, Cost measures: American, 1998, 279(6): 104-111. Practices, Part II Money, Labor, Quality, best and worst practices 3 9/18 Software Capers Jones, Conflict And Litigation Between Development Best Software Clients And Developers Practices, Part III 4 9/25 Peer Reviews of Perform an inspection exercise in class. Tom Gilb, and Dorothy Graham, Software Out: software Software and Inspection, Chapter 3, Overview of Software inspection – Other Objects Inspection what to inspect? Create teams? 5 10/2 Data Quality Data Ownership and Data Roles, Cost Analysis of Poor Data Quality, Dimensions Loshin, parts TBD Introduction of Data Quality, Data models, Data values, Data Analysis Techniques and Tools, Data Quality Improvement, Metadata and Enterprise Reference Data, Domains and Mappings Data Quality Rules—Definition and Discovery, Data Profiling, Data Transformation and Cleansing: Standardization, Linkage Duplicate Elimination, and Approximate Searching. 6 10/9 Data Quality in Data models and databases. Data flow. Costs of data defects. The information Databases, Cost of chain. Domain constraints. Integrity constraints. Low Data Quality & Quality of Data Models; Quality of Data Values; Data profiling Dimensions of Data
- 2. Quality 7 10/16 Software testing TBD TBD In: Inspection reports; Out: Data profiling 8 10/23 Bob Fitterman: Guest lecture. Fitterman is Chief Technology Officer of Vindigo, the leading www.extremeprogramming.org/rules.html and all supplier of localized information to handhelds and cell phone. He’s responsible for pages one link from it, including continued pages. Extreme supervising the design of Vindigo’s data management and synchronization http://www.vindigo.com/about/index.html and http:// programming: www.vindigo.com/learn_more.html. Vindigo’s systems. As Bob will discuss, Vindigo develops under the Extreme programming philosophy. Experience In essence, Extreme programming dramatically compresses the design-build-test- deploy cycle. It prescribes: first plan the tests, program in pairs, and release constantly. www.extremeprogramming.org/rules.html presents the details. 9 10/30 Data Types of data errors: transcription, typing, auditory, etc. Standardization 10 11/6 Ken Estes: Guest lecture. Estes is an accomplished software engineer with extensive Ken recommend links Software experience in the engineering and debugging of in-house and third-party Development applications. He is widely trained and read in Software Engineering theory and Disasters and practices, with real world experience in their application. He’s author of the current version of Tinderbox, the automated built/test monitoring Software used by Weinberg’s Views Netscape. Estes is designer and author of the run and build time dependency on Software tracking tools in the RedHat 7.0 Package Management system. Development Estes will recount software development and deployment experiences he has had while working at high tech companies and financial institutions. In addition, he will present Gerald Weinberg’s software management philosophy. Weinberg is a leading thinker on the psychology of software development, and the author of several dozen books in the area, including the 4 volume Quality Software Management series. Weinberg incorporates the precepts of family psychology, especially the work of Virginia Satir, into software project organization and management. 11 11/13 Project Effort Delphi estimation; Construx estimation technique; Perform a Wideband Delphi Karl E. Wiegers, Stop Promising Miracles In: Data estimation exercise in class. http://www.processimpact.com/articles/delphi.html profiling Estimation Generic Delphi Estimation Process http://www.cs.uwf.edu/~wilde/gump/delphi.htm http://www.construx.com/estimate/ 12 11/20 Dr. Ram Guest lecture. Chillarege was a computer scientist at IBM Research and CTO of Ram recommend a paper Chillarege: Opus360. He invented orthogonal defect classification (ODC) at IBM in the early Orthogonal Defect 90s. ODC is based on the observation that defects can be classified by type, such Classification as design, I/O, formatting, initialization, timing, etc. Defect classification measurements during a software development can accurately indicate the appropriate current development stage (design, coding, unit test, etc.) which can be contrasted to the purported stage.
- 3. 13 11/27 Data Matching: Data Matching (or Linkage): Linkage matches multiple records that correspond to Traditional and the same real entity, such as {Arthur P. Goldberg, 333 3rd Ave., # 12S, 10010, 212 Machine Learning 685-1234} with {Art Golberg, 333 Third Avenue, Apt 12 South, 10001, Approaches 212.686.1234}. We present linkage techniques, focusing on the statistical Maximum Entropy machine learning method commercialized by ChoiceMaker Technologies. 14 12/4 Ilya Pevzner guest Merging merges multiple matching records into a single record. Merging resolves Out: take home lecture. Data conflicting values, i.e. by deciding that the records above should be merged to final, due in 1 Merging: A {Arthur P. Goldberg, 333 Third Ave., Apt. 12S, 10010, ?}. Ilya will present his PhD week Machine Learning research. Approaches