Data Science Achievements#

In this course there are 5 learning outcomes that I expect you to achieve by the end of the semester. To get there, you’ll focus on 15 smaller achievements that will be the basis of your grade. This section will describe how the topics covered, the learning outcomes, and the achievements are covered over time. In the next section, you’ll see how these achievements turn into grades.

Learning Outcomes#

By the end of the semester

  1. (process) Describe the process of data science, define each phase, and identify standard tools

  2. (data) Access and combine data in multiple formats for analysis

  3. (exploratory) Perform exploratory data analyses including descriptive statistics and visualization

  4. (modeling) Select models for data by applying and evaluating mutiple models to a single dataset

  5. (communicate) Communicate solutions to problems with data in common industry formats

We will build your skill in the process and communicate outcomes over the whole semester. The middle three skills will correspond roughly to the content taught for each of the first three portfolio checks.

Schedule#

The course will meet in . Every class will include participatory live coding (instructor types code while explaining, students follow along) instruction and small exercises for you to progress toward level 1 achievements of the new skills introduced in class that day.

Each Assignment will have a deadline posted on the assignment page, typically the same day each week. Portfolio deadlines will be announced at least 2 weeks in advance.

topics skills
week
1 [admin, python review] process
2 Loading data, Python review [access, prepare, summarize]
3 Exploratory Data Analysis [summarize, visualize]
4 Data Cleaning [prepare, summarize, visualize]
5 Databases, Merging DataFrames [access, construct, summarize]
6 Modeling, classification performance metrics, cross validation [evaluate]
7 Naive Bayes, decision trees [classification, evaluate]
8 Regression [regression, evaluate]
9 Clustering [clustering, evaluate]
10 SVM, parameter tuning [optimize, tools]
11 KNN, Model comparison [compare, tools]
12 Text Analysis [unstructured]
13 Images Analysis [unstructured, tools]
14 Deep Learning [tools, compare]

Achievement Definitions#

The table below describes how your participation, assignments, and portfolios will be assessed to earn each achievement. The keyword for each skill is a short name that will be used to refer to skills throughout the course materials; the full description of the skill is in this table.

skill Level 1 Level 2 Level 3
keyword
python pythonic code writing python code that mostly runs, occasional pep8 adherance python code that reliably runs, frequent pep8 adherance reliable, efficient, pythonic code that consistently adheres to pep8
process describe data science as a process Identify basic components of data science Describe and define each stage of the data science process Compare different ways that data science can facilitate decision making
access access data in multiple formats load data from at least one format; identify the most common data formats Load data for processing from the most common formats; Compare and constrast most common formats access data from both common and uncommon formats and identify best practices for formats in different contexts
construct construct datasets from multiple sources identify what should happen to merge datasets or when they can be merged apply basic merges merge data that is not automatically aligned
summarize Summarize and describe data Describe the shape and structure of a dataset in basic terms compute summary statndard statistics of a whole dataset and grouped data Compute and interpret various summary statistics of subsets of data
visualize Visualize data identify plot types, generate basic plots from pandas generate multiple plot types with complete labeling with pandas and seaborn generate complex plots with pandas and plotting libraries and customize with matplotlib or additional parameters
prepare prepare data for analysis identify if data is or is not ready for analysis, potential problems with data apply data reshaping, cleaning, and filtering as directed apply data reshaping, cleaning, and filtering manipulations reliably and correctly by assessing data as received
evaluate Evaluate model performance Explain and compute basic performance metrics for different data science tasks Apply and interpret basic model evaluation metrics to a held out test set Evaluate a model with multiple metrics and cross validation
classification Apply classification identify and describe what classification is, apply pre-fit classification models fit, apply, and interpret preselected classification model to a dataset fit and apply classification models and select appropriate classification models for different contexts
regression Apply Regression identify what data that can be used for regression looks like fit and interpret linear regression models fit and explain regrularized or nonlinear regression
clustering Clustering describe what clustering is apply basic clustering apply multiple clustering techniques, and interpret results
optimize Optimize model parameters Identify when model parameters need to be optimized Optimize basic model parameters such as model order Select optimal parameters based of mutiple quanttiateve criteria and automate parameter tuning
compare compare models Qualitatively compare model classes Compare model classes in specific terms and fit models in terms of traditional model performance metrics Evaluate tradeoffs between different model comparison types
representation Choose representations and transform data Identify options for representing text and categorical data in many contexts Apply at least one representation to transform unstructured or inappropriately data for model fitting or summarizing apply transformations in different contexts OR compare and contrast multiple representations a single type of data in terms of model performance
workflow use industry standard data science tools and workflows to solve data science problems Solve well strucutred fully specified problems with a single tool pipeline Solve well-strucutred, open-ended problems, apply common structure to learn new features of standard tools Independently scope and solve realistic data science problems OR independently learn releated tools and describe strengths and weakensses of common tools

Assignments and Skills#

Using the keywords from the table above, this table shows which assignments you will be able to demonstrate which skills and the total number of assignments that assess each skill. This is the number of opportunities you have to earn Level 2 and still preserve 2 chances to earn Level 3 for each skill.

A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 # Assignments
keyword
python 1 1 0 1 1 0 0 0 0 0 0 0 0 4
process 1 0 0 0 0 1 1 1 1 1 1 0 0 7
access 0 1 1 1 1 0 0 0 0 0 0 0 0 4
construct 0 0 0 0 1 0 1 1 0 0 0 0 0 3
summarize 0 0 1 1 1 1 1 1 1 1 1 1 1 11
visualize 0 0 1 1 0 1 1 1 1 1 1 1 1 10
prepare 0 0 0 1 1 0 0 0 0 0 0 0 0 2
evaluate 0 0 0 0 0 1 1 1 0 1 1 0 0 5
classification 0 0 0 0 0 0 1 0 0 1 0 0 0 2
regression 0 0 0 0 0 0 0 1 0 0 1 0 0 2
clustering 0 0 0 0 0 0 0 0 1 0 1 0 0 2
optimize 0 0 0 0 0 0 0 0 0 1 1 0 0 2
compare 0 0 0 0 0 0 0 0 0 0 1 0 1 2
representation 0 0 0 0 0 0 0 0 0 0 0 1 1 2
workflow 0 0 0 0 0 0 0 0 0 1 1 1 1 4

Warning

process achievements are accumulated a little slower. Prior to portfolio check 1, only level 1 can be earned. Portfolio check 1 is the first chance to earn level 2 for process, then level 3 can be earned on portfolio check 2 or later.

Portfolios and Skills#

The objective of your portfolio submissions is to earn Level 3 achievements. The following table shows what Level 3 looks like for each skill and identifies which portfolio submissions you can earn that Level 3 in that skill.

Level 3 P1 P2 P3 P4
keyword
python reliable, efficient, pythonic code that consistently adheres to pep8 1 1 0 1
process Compare different ways that data science can facilitate decision making 0 1 1 1
access access data from both common and uncommon formats and identify best practices for formats in different contexts 1 1 0 1
construct merge data that is not automatically aligned 1 1 0 1
summarize Compute and interpret various summary statistics of subsets of data 1 1 0 1
visualize generate complex plots with pandas and plotting libraries and customize with matplotlib or additional parameters 1 1 0 1
prepare apply data reshaping, cleaning, and filtering manipulations reliably and correctly by assessing data as received 1 1 0 1
evaluate Evaluate a model with multiple metrics and cross validation 0 1 1 1
classification fit and apply classification models and select appropriate classification models for different contexts 0 1 1 1
regression fit and explain regrularized or nonlinear regression 0 1 1 1
clustering apply multiple clustering techniques, and interpret results 0 1 1 1
optimize Select optimal parameters based of mutiple quanttiateve criteria and automate parameter tuning 0 0 1 1
compare Evaluate tradeoffs between different model comparison types 0 0 1 1
representation apply transformations in different contexts OR compare and contrast multiple representations a single type of data in terms of model performance 0 0 1 1
workflow Independently scope and solve realistic data science problems OR independently learn releated tools and describe strengths and weakensses of common tools 0 0 1 1

Detailed Checklists#

python-level1#

python code that mostly runs, occasional pep8 adherance

  • [ ] use of control structures

  • [ ] callable functions

  • [ ] correct calls to functions

  • [ ] correct use of variables

  • [ ] use of logical operators

python-level2#

python code that reliably runs, frequent pep8 adherance

  • [ ] descriptive variable names

  • [ ] pythonic loops

  • [ ] effective use of return vs side effects in functions

  • [ ] correct, effective use of builtin python iterable types (lists & dictionaries)

python-level3#

reliable, efficient, pythonic code that consistently adheres to pep8

  • [ ] pep8 adherant variable, file, class, and function names

  • [ ] effective use of multi-paradigm abilities for efficiency gains

  • [ ] easy to read code that adheres to readability over other rules

process-level1#

Identify basic components of data science

  • [ ] identify component disciplines

  • [ ] idenitfy phases

process-level2#

Describe and define each stage of the data science process

  • [ ] correctly defines stages

  • [ ] identifies stages in use

  • [ ] describes general goals as well as a specific processes

process-level3#

Compare different ways that data science can facilitate decision making

  • [ ] describes exceptions to process and iteration in process

  • [ ] connects choices at one phase to impacts in other phases

  • [ ] connects data science steps to real world decisions

access-level1#

load data from at least one format; identify the most common data formats

  • [ ] use at least one pandas read_ function correctly

  • [ ] name common types

  • [ ] describe the structure of common types

access-level2#

Load data for processing from the most common formats; Compare and constrast most common formats

  • [ ] load data from at least two of (.csv, .tsv, .dat, database, .json)

  • [ ] describe advantages and disadvantages of most commone types

  • [ ] descive how most common types are different

access-level3#

access data from both common and uncommon formats and identify best practices for formats in different contexts

  • [ ] load data from at least 1 uncommon format

  • [ ] describe when one format is better than another

construct-level1#

identify what should happen to merge datasets or when they can be merged

  • [ ] identify what the structure of a merged dataset should be (size, shape, columns)

  • [ ] idenitfy when datasets can or cannot be merged

construct-level2#

apply basic merges

  • [ ] use 3 different types of merges

  • [ ] choose the right type of merge for realistic scenarios

construct-level3#

merge data that is not automatically aligned

  • [ ] manipulate data to make it mergable

  • [ ] identify how to combine data from many sources to answer a question

  • [ ] implement stesp to combine data from multiple sources

summarize-level1#

Describe the shape and structure of a dataset in basic terms

  • [ ] use attributes to produce a description of a dataset

  • [ ] display parts of a dataset

summarize-level2#

compute and interpret summary standard statistics of a whole dataset and grouped data

  • [ ] compute descriptive statistics on whole datasets

  • [ ] apply individual statistics to datasets

  • [ ] group data by a categorical variable for analysis

  • [ ] apply split-apply-combine paradigm to analyze data

  • [ ] interprete statistics on whole datasets

  • [ ] interpret statistics on subsets of data

summarize-level3#

Compute and interpret various summary statistics of subsets of data

  • [ ] produce custom aggregation tables to summarize datasets

  • [ ] compute multivariate summary statistics by grouping

  • [ ] compute custom cacluations on datasets

visualize-level1#

identify plot types, generate basic plots from pandas

  • [ ] generate at least two types of plots with pandas

  • [ ] identify plot types by name

  • [ ] interpret basic information from plots

visualize-level2#

generate multiple plot types with complete labeling with pandas and seaborn

  • [ ] generate at least 3 types of plots

  • [ ] use correct, complete, legible labeling on plots

  • [ ] plot using both pandas and seaborn

  • [ ] interpret multiple types of plots to draw conclusions

visualize-level3#

generate complex plots with pandas and plotting libraries and customize with matplotlib or additional parameters

  • [ ] use at least two libraries to plot

  • [ ] generate figures with subplots

  • [ ] customize the display of a plot to be publication ready

  • [ ] interpret plot types and explain them for novices

  • [ ] choose appopriate plot types to convey information

  • [ ] explain why plotting common best practices are effective

prepare-level1#

identify if data is or is not ready for analysis, potential problems with data

  • [ ] identify problems in a dataset

  • [ ] anticipate how potential data setups will interfere with analysis

  • [ ] describe the structure of tidy data

  • [ ] label data as tidy or not

prepare-level2#

apply data reshaping, cleaning, and filtering as directed

  • [ ] reshape data to be analyzable as directed

  • [ ] filter data as directed

  • [ ] rename columns as directed

  • [ ] rename values to make data more analyzable

  • [ ] handle missing values in at least two ways

  • [ ] transform data to tidy format

prepare-level3#

apply data reshaping, cleaning, and filtering manipulations reliably and correctly by assessing data as received

  • [ ] identify issues in a dataset and correctly implement solutions

  • [ ] convert varialbe representation by changing types

  • [ ] change variable representation using one hot encoding

evaluate-level1#

Explain and compute basic performance metrics for different data science tasks

  • [ ] apply at least one metric

  • [ ] interpret model performance in context

evaluate-level2#

Apply and interpret basic model evaluation metrics to a held out test set

  • [ ] apply at least three performance metrics to models

  • [ ] apply metrics to subsets of data

  • [ ] apply disparity metrics

  • [ ] interpret at least three metrics

evaluate-level3#

Evaluate a model with multiple metrics and cross validation

  • [ ] explain cross validation

  • [ ] explain importance of held out test and validation data

  • [ ] describe why cross vaidation is important

  • [ ] idenitfy appropriate metrics for different types of modeling tasks

  • [ ] use multiple metrics together to create a more complete description of a model’s performance

classification-level1#

identify and describe what classification is, apply pre-fit classification models

  • [ ] describe what classification is

  • [ ] describe what a dataset must look like for classifcation

  • [ ] identify appliations of classifcation in the real world

  • [ ] describe set up for a classification problem (tes,train)

classification-level2#

fit, apply, and interpret preselected classification model to a dataset

  • [ ] split data for training and testing

  • [ ] fit a classification model

  • [ ] apply a classification model to obtain predictions

  • [ ] interpret the predictions of a classification model

  • [ ] examine parameters of at least one fit classifier to explain how the prediction is made

  • [ ] differentiate between model fitting and generating predictions

  • [ ] evaluate how model parameters impact model performance

classification-level3#

fit and apply classification models and select appropriate classification models for different contexts

  • [ ] choose appropriate classifiers based on application context

  • [ ] explain how at least 3 different classifiers make predictions

  • [ ] evaluate how model parameters impact model performance and justify choices when tradeoffs are necessary

regression-level1#

identify what data that can be used for regression looks like

  • [ ] identify data that is/not appropriate for regression

  • [ ] describe univariate linear regression

  • [ ] identify appliations of regression in the real world

regression-level2#

fit and interpret linear regression models

  • [ ] split data for training and testing

  • [ ] fit univariate linear regression models

  • [ ] interpret linear regression models

  • [ ] fit multivariate linear regression models

regression-level3#

fit and explain regrularized or nonlinear regression

  • [ ] fit nonlinear or regrularized regression models

  • [ ] interpret and explain nonlinear or regrularized regresion models

clustering-level1#

describe what clustering is

  • [ ] differentiate clustering from classification and regression

  • [ ] identify appliations of clustering in the real world

clustering-level2#

apply basic clustering

  • [ ] fit Kmeans

  • [ ] interpret kmeans

  • [ ] evaluate clustering models

clustering-level3#

apply multiple clustering techniques, and interpret results

  • [ ] apply at least two clustering techniques

  • [ ] explain the differences between two clustering models

optimize-level1#

Identify when model parameters need to be optimized

  • [ ] identify when parameters might impact model performance

optimize-level2#

Optimize basic model parameters such as model order

  • [ ] automatically optimize multiple parameters

  • [ ] evaluate potential tradeoffs

  • [ ] interpret optimization results in context

optimize-level3#

Select optimal parameters based of mutiple quanttiateve criteria and automate parameter tuning

  • [ ] optimize models based on multiple metrics

  • [ ] describe when one model vs another is most appropriate

compare-level1#

Qualitatively compare model classes

  • [ ] compare models within the same task on complexity

compare-level2#

Compare model classes in specific terms and fit models in terms of traditional model performance metrics

  • [ ] compare models in multiple terms

  • [ ] interpret cross model comparisons in context

compare-level3#

Evaluate tradeoffs between different model comparison types

  • [ ] compare models on multiple criteria

  • [ ] compare optimized models

  • [ ] jointly interpret optimization result and compare models

  • [ ] compare models on quanttiateve and qualitative measures

representation-level1#

Identify options for representing text and categorical data in many contexts

  • [ ] describe the basic goals for changing the representation of data

representation-level2#

Apply at least one representation to transform unstructured or inappropriately data for model fitting or summarizing

  • [ ] transform text or image data for use with ML

representation-level3#

apply transformations in different contexts OR compare and contrast multiple representations a single type of data in terms of model performance

  • [ ] transform both text and image data for use in ml

  • [ ] evaluate the impact of representation on model performance

workflow-level1#

Solve well strucutred fully specified problems with a single tool pipeline

  • [ ] pseudocode out the steps to answer basic data science questions

workflow-level2#

Solve well-strucutred, open-ended problems, apply common structure to learn new features of standard tools

  • [ ] plan and execute answering real questions to an open ended question

  • [ ] describe the necessary steps and tools

workflow-level3#

Independently scope and solve realistic data science problems OR independently learn releated tools and describe strengths and weakensses of common tools

  • [ ] scope and solve realistic data science problems

  • [ ] compare different data science tool stacks