Your Life on a Map: Private Geospatial Analytics with Kotlin DataFrame and Kandy

Demo Repository: https://github.com/zaleslaw/Geo-Analytics-KotlinConf-2026

Transcript

1. Your Life on a Map: Private Geospatial Analytics with Kotlin DataFrame and Kandy Aleksei Zinovev Team Lead in Kotlin for Data Science
2. About • Team Lead, Kotlin for Data Science • Geographical Data Fan • Math Enthusiast • Kotlin DataFrame Contributor • KotlinDL Author • Apache Ignite (in-memory database) PMC
4. Why the App Is Not Enough • Single activity view only
5. Why the App Is Not Enough • Single activity view only • No global map across activities
6. Why the App Is Not Enough • Single activity view only • No global map across activities • Closed data storage
7. Why the App Is Not Enough • Single activity view only • No global map across activities • Closed data storage • Export one track at a time
8. Why the App Is Not Enough • Single activity view only • No global map across activities • Closed data storage • Export one track at a time
9. What I Need Instead
10. Drop Your First geoTrack
11. What GPX Looks Like <trkpt lat="52.552857" lon="13.424176"> <time>2026-03-12T09:19:27.054Z</time> </trkpt> <trkpt lat="52.552812" lon="13.424094"> <time>2026-03-12T09:19:28.500Z</time> </trkpt>
12. GPX → Kotlin val points = parseGPX(filePath)
13. GPX → Kotlin val points = parseGPX(filePath) data class TrackPoint( val lat: Double, val lon: Double, val time: String, val altitude: Double? = null )
14. DataFrame?
15. Read
16. Read → Transform
17. Read → Transform → Export
18. GPX → DataFrame val df = DataFrame.readGPX("walk1.gpx")
19. GPX → DataFrame fun DataFrame.Companion.readGPX(filePath: String): DataFrame<*> { val points = parseGPX(filePath) return dataFrameOf( "lat" to points.map { it.lat }, "lon" to points.map { it.lon }, "time" to points.map { it.time }, "altitude" to points.map { it.altitude } )
20. GPX → DataFrame val df = DataFrame.readGPX("walk1.gpx")
21. Just One Random Track Visualized With Kandy
22. GPX → Kandy Plot val df = DataFrame.readGPX("walk1.gpx")
23. GPX → Kandy Plot val df = DataFrame.readGPX("walk1.gpx") val geoDataFrame = gpxDf.toGeoDataFrame()
24. GPX → Kandy Plot val df = DataFrame.readGPX("walk1.gpx") val geoDataFrame = gpxDf.toGeoDataFrame() geoDataFrame.plot { geoPoints { color = Color.ORANGE } }
25. Just One Random Track Visualized With Kandy
26. Just One Random Track Visualized With Kandy
27. Just One Random Track Visualized With Kandy
28. GPX → Kandy Plot (Points + Polygons) val berlinBezirke = GeoDataFrame.readShapefile("Berlin_Bezirke.shp")
29. GPX → Kandy Plot (Points + Polygons) val berlinBezirke = GeoDataFrame.readShapefile("Berlin_Bezirke.shp")
30. GPX → Kandy Plot (Points + Polygons) berlinBezirke.plot { geoPolygon() withData(geoDataFrame) { geoPoints { size = 1.5 } } }
31. Join 300+ Tracks together
32. All My Berlin Tracks
33. All My Berlin Tracks
34. All My Berlin Tracks
36. Temporal Patterns Discovery
37. Data Enrichment Pipeline val enrichedTracks = tracks.addTemporalAttributes() .addDistanceColumns() .addSpeed()
38. Duration By Hour
39. Duration By Month
40. Scaled Colored Bars With Kandy. Part I val byMonth = this.add("dur_hours") {(this["duration_minutes"] as Long).toDouble() / 60.0 } .groupBy { "month"<Int>() } .aggregate { sum { "dur_hours"<Double>() } into "total_hours" } .sortBy("month") .add("season") { when (this["month"] as Int) { 12, 1, 2 -> "Winter" 3, 4, 5 -> "Spring" 6, 7, 8 -> "Summer" else -> "Fall" }
41. Scaled Colored Bars With Kandy. Part II bars { x("month") { axis.name = "Month" } y("total_hours") { axis.name = "Total Hours" } alpha = 0.85 fillColor("season") { scale = categorical( "Winter" to Color.hex("#5B9BD5"), "Spring" to Color.hex("#70AD47"), "Summer" to Color.hex("#FFC000"), "Fall" to Color.hex("#ED7D31") )
42. Scaled Colored Bars With Kandy. Part III layout { title = "Walk Duration by Month (hours)" size = 1200 to 800 style { axis { text { fontSize = 25.0 } title { fontSize = 27.0 } } legend { text { fontSize = 25.0 } title { fontSize = 27.0 }
43. Clean the data!
44. Raw Speeds
45. Why GPS Data Looks Broken
46. Cleaning Strategy val smoothedDf = combinedDf .aggregatePointsByTime(intervalSeconds = 10) What happens inside: 1. Groups raw points into 10-sec windows
47. Cleaning Strategy val smoothedDf = combinedDf .aggregatePointsByTime(intervalSeconds = 10) What happens inside: 1. Groups raw points into 10-sec windows 2. Averages lat/lon → recalculates speed
48. Cleaning Strategy val smoothedDf = combinedDf .aggregatePointsByTime(intervalSeconds = 10) What happens inside: 1. Groups raw points into 10-sec windows 2. Averages lat/lon → recalculates speed 3. Spike 47 km/h → clean 4–6 km/h
49. Intervals By Speed Range
50. Intervals By Speed Range After Cleanup
51. After Cleanup
52. All About The Rain
53. Weather Enrichment Service
54. Walks Per Month By Weather
55. Walks By Temperature Range
56. My Hot Points in Berlin
57. Time For Heatmap
58. Your GPS History As A Heatmap In 4 steps • Count — split the map into a grid, count how many times you visited each cell
59. Your GPS History As A Heatmap In 4 steps • Count — split the map into a grid, count how many times you visited each cell • Blur — spread each cell's heat to its neighbors (repeat 3×) — edges become smooth
60. Your GPS History As A Heatmap In 4 steps • Count — split the map into a grid, count how many times you visited each cell • Blur — spread each cell's heat to its neighbors (repeat 3×) — edges become smooth • Log scale — so quiet streets don't disappear next to your favorite coffee shop
61. Your GPS History As A Heatmap In 4 steps • Count — split the map into a grid, count how many times you visited each cell • Blur — spread each cell's heat to its neighbors (repeat 3×) — edges become smooth • Log scale — so quiet streets don't disappear next to your favorite coffee shop • Paint — blue (rarely visited) → green → yellow → red (your usual routes)
62. createSmoothHeatmap Example Call walksDf.createSmoothHeatmap( shapefileDf = berlinWgs84, xRange = 13.08..13.73, yRange = 52.34..52.67, resolution = 200, blurMeters = 450.0, blurPasses = 3, )
63. Heatmaps Are Different
64. Heatmaps Are Different
65. Tech Stack Summary • Kotlin DataFrame — data processing and transformations
66. Tech Stack Summary • Kotlin DataFrame — data processing and transformations • Kandy — declarative visualization
67. Tech Stack Summary • Kotlin DataFrame — data processing and transformations • Kandy — declarative visualization • dataframe-geo + JTS — geospatial DataFrames and polygon geometry
68. Tech Stack Summary • Kotlin DataFrame — data processing and transformations • Kandy — declarative visualization • dataframe-geo + JTS — geospatial DataFrames and polygon geometry • GeoTools — coordinate system conversions (CRS / EPSG)
69. Tech Stack Summary • Kotlin DataFrame — data processing and transformations • Kandy — declarative visualization • dataframe-geo + JTS — geospatial DataFrames and polygon geometry • GeoTools — coordinate system conversions (CRS / EPSG) • Ktor Client — HTTP calls to Open-Meteo weather API
70. Tech Stack Summary • Kotlin DataFrame — data processing and transformations • Kandy — declarative visualization • dataframe-geo + JTS — geospatial DataFrames and polygon geometry • GeoTools — coordinate system conversions (CRS / EPSG) • Ktor Client — HTTP calls to Open-Meteo weather API • kotlinx-datetime — temporal operations
71. Try It Yourself • Export your GPS data (GPX, KML, TCX formats) • Vibecode YourFileFormat to DataFrame read function • Use Kotlin DataFrame for processing • Enrich with Open-Meteo for weather • Visualize with Kandy
72. Links • Private Geo Analytics Tutorial • Kandy Geo Tutorial • Geo Examples • DataFrame Docs • Data Analysis on Kotlin Site
73. github.com/zaleslaw/Geo-Analytics-KotlinConf-2026