swaroop-p / machine-learning-1

Solutions to machine-learning problems on Hackerearth and Kaggle

Machine-learning

Contains solutions to machine-learning problems hosted on various sites like Hackerearth

RoadSign Prediction :

Predict which direction a road sign is applicable to given the following features :

1. ID - a unique identifier for each record in dataset.
2. DetectedCamera - imagine a car, which is fitted with 4 cameras on top, one each facing front, right, rear and left. Each cameras clicks pictures on every few meters as car moves. DetectedCamera value tells you on which camera image a road sign was observed/found, by an image detection software.
3. AngleOfSign - values are in degrees ranging from 0 to 360 in clockwise direction from the front of car, indicates the angle from the front of the car to the direction in which the sign is detected.
4. SignWidth - width of the sign bounding box in the image in pixels.
5. SignHeight - height of the sign bounding box in the image in pixels.
6. SignAspectRatio - this is the width/height ratio of the sign bounding box, derived from SignWidth/SignHeight. Can provide an indication that sign is facing camera or not. A sign facing the driver, detected on an image captured from almost 80 degrees from front (on right camera), will have a bounding box that is skewed from its original aspect ratio. If its facing the right camera, it will have nearly original aspect ratio of the sign.
7. SignFacing (Target) - For the above inputs, where the sign is actually facing is captured here, from manually reviewed sign facing records.

I finished under top 15 in this competetion. There was a dataleak i missed. Row ids closer to each other usually belonged to same class.

Predict the Segment - Hotstar :

classify customers based on watch patterns, learn patterns from customers whose watch patterns are already known.

1. ID - unique identifier variable.
2. titles - titles of the shows watched by the user and watch_time on different titles in the format “title:watch_time” separated by comma, e.g. “JOLLY LLB:23, Ishqbaaz:40”. watch_time is in seconds
3. genres - same format as titles.
4. cities - same format as titles.
5. tod - total watch time of the user spreaded across different time of days (24 hours format) in the format “time_of_day:watch_time” separated by comma, e.g. “1:454, “17”:5444”.
6. dow - total watch time of the user spreaded across different days of week (7 days format) in the format “day_of_week:watch_time” separated by comma, e.g. “1:454, “6”:5444”.
7. segment - target variable. consider them as interest segments. For modeling, encode pos = 1, neg = 0.

Finished in the top 10% and qualified to the second round after combining both road sign and segment prediction scores. Using an lgbm model would got the highest score on the leader board.

Predict Ad-clicks :

Predict the probability whether an ad will get clicked or not.

1. ID - Unique ID
2. datetime - timestamp
3. siteid - website id
4. offerid - offer id (commission based offers)
5. category - offer category
6. merchant - seller ID
7. countrycode - country where affiliates reach is present
8. browserid - browser used
9. devid - device used
10. click - target variable

Predict Happiness based on Hotel reviews :

Text mininng to predict customer happiness.

Whats cooking ?

Predict cuisine based on ingredients.