BDB650

Kanban and CRISP-DM

Summary

Kanban Overview

Kanban Boards

CRISP-DM Overview

CRISP-DM - Phases

Kanban

Kanban's History

Developed at Toyota in the 1940's

Aligned inventory levels with the actual consumption of materials

Requires the passing of cards (kanban)

Uses a Just In Time (JIT) approach

Kanban in Agile

Kanban is a popular Framework for Agile development

It matches the amount of work in progress (WIP) to the team's capacity

Allows for flexible planning, fast output, and clear focus

Kanban Boards

Kanban Boards

Used to visualize tasks and optimize the workflow

Default: To-Do, In Progress, and Done

More columns can be added: testing, waiting, etc.

Provides a single source of truth for the team

Kanban Cards

Feature critical info about work items:

Description of the task

Who is reponsible

Time estimates

etc.

Kanban System

No specific timelines for sprints

Each column of the board has a limited capacity - WIP

As cards move columns, more cards are pulled from the backlog

Scrum vs Kanban

Scrum Kanban
Cadence Fixed sprints continuous flow
Change no change mid-sprint at any point
Roles PO, SM, and Dev Team No specific roles

CRISP-DM

Overview

CRoss-Industry Standard Process for Data Mining

First published in 1999 by an European consortium

Still widely used in many companies, such as IBM

Overview

Breaks down Data Mining Projects into six phases:

Business Understanding

Data Understanding

Data Preparation

Modelling

Evaluation

Deployment

Broken down in Phases

Hierarchical Method

Phase: Six in total

Generic Task: breakdown of phases

Specialized Task: how actions in the generic tasks should be carried out

Process Instance: record of the actions, decisions, and results

Hierarchy

Example of Hierarchy

Phase: Data Preparation

Generic Task: Clean Data

Specialized Task: Remove erroneous values

Process Instance: Identify erroneous data, decide on a strategy for removing erroneus values, implement strategy, document the process

CRISP-DM Phases

Business Understanding

Understanding of objectives and requirements

Assessing the situation

Creating a plan of action

Four Tasks

Data Understanding

Identifying, collecting, and analyzing data sets

This also include verifying the quality of the data

Four Tasks

Data Preparation

Preparing data sets for modeling

AKA: "data wrangling" or "data munging"

Often, amounts to about 80% of the project effort

Five Tasks

Modelling

Most exciting, but generally the shortest phase of the project

Building and assessing various models

Ideally, an iterative process

Four Tasks

Evaluation

Looking at which models best meets the business needs

This contrasts with the previous, more technical assessment in modelling

Three Tasks

Deployment

Delivering the results:

  sharing a report

  presenting the findings

  implementing a data mining process

Four Tasks

Reading

Kanban

Kanban vs Scrum (required)

Kanban WIP

CRISP-DM

CRISP-DM Guide