• Like
HunchLab 2.0 Predictive Missions: Under the Hood
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

HunchLab 2.0 Predictive Missions: Under the Hood

  • 658 views
Published

 

Published in Technology , Education
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
658
On SlideShare
0
From Embeds
0
Number of Embeds
2

Actions

Shares
Downloads
6
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. Missions: Under the Hood 340 N 12th St, Suite 402 Philadelphia, PA 19107 215.925.2600 info@azavea.com www.hunchlab.com
  • 2. Jeremy Heffner HunchLab Product Manager jheffner@azavea.com 215.701.7712 Amelia Longo Business Development Associate alongo@azavea.com 215.701.7715
  • 3. Places People Patterns } Prioritization
  • 4. Predictive Missions
  • 5. It’s the fourth Tuesday in January and school is in session. There were 3 burglaries and 2 robberies yesterday. Six bars, three take-out stores, and a school are in the neighborhood. The forecast is 17° with cloudy skies. Where do you focus your 2 vehicles?
  • 6. How would you do it?
  • 7. Analyst Process •  Identify relevant factors –  Training / Literature –  Experience •  Use heuristics –  –  –  –  –  high concentration of past crime è higher risk near a bar on a Friday night è higher risk near the police station è lower risk concentration of ex-offenders è higher risk near transit stops è higher risk
  • 8. ?  
  • 9. How HunchLab Works
  • 10. term: machine learning A computer system designed to learn how to accomplish a task by using historic data sets. There are different ways (algorithms) to accomplish this training process.
  • 11. term: algorithm The step-by-step procedure to accomplish a given calculation. Different algorithms have different qualities. Algorithms are used to train a machine learning model.
  • 12. Overall Process 1.  Generate training examples of outcomes 2.  Enrich with relevant variables 3.  Build models 4.  Evaluate accuracy 5.  Select best performing model
  • 13. Generate Examples
  • 14. ~ 500 ft cells & 1+ hour time slices
  • 15. Data Volume •  Space –  Lincoln, NE is 90 sq miles –  500 ft cell size creates 12,000 cells •  Time –  3 years of data –  1 hour resolution –  26,000 hour blocks •  Space x Time –  312,000,000 hour block cells (examples)
  • 16. Data Volume •  Space –  Lincoln, NE is 90 sq miles –  500 ft cell size creates 12,000 cells •  Time –  3 years of data –  1 hour resolution –  26,000 hour blocks •  Space x Time –  312,000,000 hour block cells (examples) •  Sampling FTW! –  Outcomes are sparse (small % of examples have crimes) –  Sampling strategy preserves crime events
  • 17. Representing Crime Theories
  • 18. Predictive Missions •  Crime predictions based on: –  Baseline crime levels •  Similar to traditional hotspot maps –  Near repeat patterns •  Event recency (contagion) –  Risk Terrain Modeling •  Proximity and density of geographic features •  Points, Lines, Polygons (bars, bus stops, etc.) –  Collective Efficacy •  Socioeconomic indicators (poverty, unemployment, etc.)
  • 19. Predictive Missions •  Crime predictions based on: –  Routine Activity Theory •  Offender: proximity and concentration of known offenders •  Guardianship: police presence (AVL / GPS) •  Targets: measures of exposure (population, parcels, vehicles) –  Temporal cycles •  Seasonality, time of month, day of week, time of day –  Recurring temporal events •  Holidays, sporting events, etc. –  Weather •  Temperature, precipitation
  • 20. Representing Crime Theories Risk Terrain Modeling
  • 21. Gun  shoo)ngs  example   Source:  Rutgers,  h8p://www.rutgerscps.org/rtm/irvrtmgoogearth.htm  
  • 22. crimes prior7 prior364 dayssincelast bardist dow 0 0 0 365 >2000ft Monday 0 0 1 234 >2000ft Monday 1 1 3 3 750ft Tuesday 0 0 2 43 500ft Wednesday 2 0 2 74 500ft Friday
  • 23. Representing Crime Theories Aoristic Analysis
  • 24. crimes probability 0 0 1 a 2 b 3 c 4 d
  • 25. crimes weights prior7 prior364 0 1 0 0 1 0 dayssincelast bardist dow 0 365 >2000ft Monday 0 1 234 >2000ft Monday 0.5 1 3 3 750ft Tuesday 1 0.5 1 3 3 750ft Tuesday 0 0 0 2 43 500ft Wednesday 0 0.13 0 2 74 500ft Friday 1 0.32 0 2 74 500ft Friday 2 0.55 0 2 74 500ft Friday
  • 26. Building Models
  • 27. Models •  Baseline –  Baseline models (6) •  Counts –  28 day –  56 day –  364 day •  Kernel Densities –  28 day –  56 day –  364 day –  HunchLab models •  Variations of a stacked ensemble: –  examples è gradient boosting machine (gbm) è y/n probabilities –  y/n probabilities è generalized additive model (gam) è counts
  • 28. term: decision tree A machine learning algorithm that recursively partitions a data set based upon variable values forming a tree-like structure.
  • 29. crimes prior7 prior364 dayssincelast bardist dow 0 0 0 365 >2000ft Monday 0 0 1 234 >2000ft Monday 1 1 3 3 750ft Tuesday 0 0 2 43 500ft Wednesday 2 0 2 74 500ft Friday
  • 30. term: gradient boosting machine (GBM) A machine learning algorithm that uses a series of weaker models (typically decision trees) that are trained upon the residuals of prior iterations (boosting) to form one stronger model. 1 Build Decision Tree 1 2 Build Decision Tree 2 3 Build Decision Tree 3 Predict with 1 Predict with 1 & 2 Predict with 1-3 Calculate errors Calculate errors Calculate errors …
  • 31. term: generalized additive model (GAM) A regression model that fits smoothed functions to the input variables. Compare to a generalized linear model which fits just a single coefficient to each variable.
  • 32. HunchLab Model Building 1.  Build a GBM –  examples è gradient boosting machine è y/n probabilities
  • 33. 312 million Sampling 4 million 4 folds 1 mil 1 mil 1 mil 1 mil } GBM 1 mil Evaluate 43 200
  • 34. 312 million Sampling 4 million GBM 43
  • 35. HunchLab Model Building 1.  Build a GBM –  examples è gradient boosting machine è y/n probabilities •  Segment examples into several folds –  For each fold build a GBM model on the rest of the data –  For each iteration in the GBMs: »  Randomly sample a portion of the data (stochastic) »  Adjust weights of observations (adaptive boosting) •  Determine how many iterations result in the most accurate model •  Build a GBM on all of the data for that many iterations
  • 36. HunchLab Model Building 2.  Build a GAM –  y/n probabilities è generalized additive model è counts •  Transforms (“bends”) GBM output into counts •  Calibrates count levels with other key variables
  • 37. Example
  • 38. Lincoln NE
  • 39. Lincoln Assaults
  • 40. Lincoln Assaults
  • 41. Lincoln Assaults
  • 42. Lincoln Assaults
  • 43. Lincoln Assaults
  • 44. Selecting Models
  • 45. Selecting Models 1.  Build models holding out last 28 days of data 2.  Score each model –  Combine different metrics into a selection score 3.  Select best score 4.  Rebuild the best model (including last 28 days data)
  • 46. A map represented as a grid of cells Crime Location 100% 0% Cells ranked highest to lowest
  • 47. 100% 0% Cells ranked highest to lowest Percent of Patrol Area to Capture All Crimes Average Crime Rank 100% 50% Percent of Crimes Captured vs. Percent of Patrol Area 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
  • 48. Assault   Burglary   MVT   Rape   Robbery   Percent  of  Patrol  Area  to  Capture  All  Crimes   0%   20%   40%   60%   80%   100%  
  • 49. Average  Crime  Rank   Assault 0 0.1 0.2 0.3 0.4 0.5 0.6 Burglary MVT Rape Robbery
  • 50. Theft of Motor Vehicle Percent of Crimes Captured 0.4 0.2 0 0 0.02 0.04 0.06 0.08 Percent of Land Area 0.1 0.12 0.14 0.16
  • 51. Overall Process 1.  Generate training examples of outcomes 2.  Enrich with relevant variables 3.  Build models 4.  Evaluate accuracy 5.  Select best performing model
  • 52. Our Solution •  Learns from several years of your data •  Automatically determines which theories apply –  more than just crime data •  Prevents over-fitting •  Calibrates predictions •  Selects a model based upon a blind evaluation –  prioritization and count-based metrics
  • 53. Our Solution •  Learns from several years of your data •  Automatically determines which theories apply –  more than just crime data •  Prevents over-fitting •  Calibrates predictions •  Selects a model based upon a blind evaluation –  prioritization and count-based metrics •  But it still cannot make your morning coffee
  • 54. Additional Information •  How did HunchLab originate? •  How does HunchLab represent crime theories? •  What data is needed? •  How does the modeling work specifically?
  • 55. Questions 340 N 12th St, Suite 402 Philadelphia, PA 19107 215.925.2600 info@azavea.com www.hunchlab.com
  • 56. Jeremy Heffner HunchLab Product Manager jheffner@azavea.com 215.701.7712 Amelia Longo Business Development Associate alongo@azavea.com 215.701.7715 340 N 12th St, Suite 402 Philadelphia, PA 19107 215.925.2600 info@azavea.com www.hunchlab.com