Travel Data Analysis Using spark (use case)
Travel Data Analysis Using spark (use case)
In this blog, we will discuss the analysis of travel dataset and gain insights from the dataset using Apache Spark.
The travel dataset is publically available and the contents are detailed under the heading, ‘Travel Sector Dataset Description’.
Based on the data, we will find the top 20 destination people travel the most, top 20 locations from where people travel the most, top 20 cities that generate high airline revenues for travel, based on booked trip count.
Travel Sector Dataset Description:-
Column 1: City pair (Combination of from and to): String
Column 2: From location: String
Column 3: To Location: String
Column 4: Product type: Integer (1=Air, 2=Car, 3 =Air+Car, 4 =Hotel, 5=Air+Hotel, 6=Hotel +Car, 7 =Air+Hotel+Car)
Column 5: Adults traveling: Integer
Column 6: Seniors traveling: Integer
Column 7: Children traveling: Integer
Column 8: Youth traveling: Integer
Column 9: Infant traveling: Integer
Column 10: Date of travel: String
Column 11: Time of travel: String
Column 12: Date of Return: String
Column 13: Time of Return: String
Column 14: Price of booking: Float
Column 15: Hotel name: String
You can download the dataset from the link below:-
Problem Statement 1:-
Top 20 destination people travel the most: Based on the given data, we can find the most popular destination that people travel frequently. There are many destinations out of which we will find only first 20, based on trips booked for particular destinations.
Source Code:-
val textFile = sc.textFile("hdfs://localhost:9000/TravelData.txt")
val split = textFile.map(lines=>lines.split('\t')).map(x=>(x(2),1)).reduceByKey(_+_).map(item => item.swap).sortByKey(false).take(20)
Description of the above code:-
Line 1: We are creating an RDD by loading a new dataset which is in HDFS.
Line 2: We have split each record by taking the delimiter as tab because the data is tab separated. We are creating the key-value pair, where a key is a destination that is in 3rd column and the value is 1. Since we need to count the cities which are popular, we are using the reduceByKey method to count them. After counting the destinations, we are swapping the key-value pairs. The sortByKey method sorts the data with keys and false stands for descending order. Once the sorting is complete, we are considering the top 20 destinations.
Output
(396,MIA), (290,SFO), (202,LAS), (162,LAX), (102,DFW), (64,DEN), (57,ORD), (54,PHL), (50,IAH), (45,JFK), (44,PHX), (40,FLL), (36,ATL), (31,BOS), (31,MCO), (27,SAN), (25,WAS), (24,CUN), (22,AUS), (22,LON)
You can see the same in the below screen shot.
Problem Statement 2:-
Top 20 locations from where people travel the most: We can find the places from where most of the trips are undertaken, based on the booked trip count.
Source Code:-
val textFile = sc.textFile("hdfs://localhost:9000/TravelData.txt")
val split = textFile.map(lines=>lines.split('\t')).map(x=>(x(1),1)).reduceByKey(_+_).map(item => item.swap).sortByKey(false).take(20)
Description of the above code:-
Line 1: We are creating an RDD by loading a new dataset which is in HDFS.
Line 2: We have split each record by taking the delimiter as tab since the data is tab separated. We are creating the key-value pair, where a key is a location from where people start, that is in the 2nd column and the value is 1. Since we need to count the cities which are popular locations from where people undertake the trips, we are using the reduceByKey method to count them. After counting the locations, we are swapping the key-value pairs. We are using the sortByKey method which sorts the data with keys where false stands for descending order. Once the sorting is complete, we are taking the top 20 locations from where people undertake the trips.
Output:-
(504,DFW), (293,MIA), (272,LAS), (167,BOM), (131,SFO), (101,ORD), (72,LAX), (55,DEN), (41,PHL), (37,IAH), (35,FLL), (33,PHX), (31,JFK), (24,WAS), (19,HOU), (19,ATL), (18,DXB), (17,SAN), (17,BOS), (17,BCN)
You can see the same in the below screen shot
Problem Statement 3:-
Top 20 cities that generate high airline revenues for travel, so that the site can concentrate on offering the discount on booking, to those cities to attract more bookings.
Source Code:-
val textFile = sc.textFile("hdfs://localhost:9000/TravelData.txt")
val fil = textFile.map(x=>x.split('\t')).filter(x=>{if((x(3).matches(("1")))) true else false })
val cnt = fil.map(x=>(x(2),1)).reduceByKey(_+_).map(item => item.swap).sortByKey(false).take(20)
Description of the above code:-
Line 1: We are creating an RDD by loading a new dataset which is in HDFS.
Line 2: We are splitting each record based on the delimiter tab as the data is tab separated. From this, we are filtering the records based on the mode of travel. Here, we need the count of people who traveled by flight which is denoted by 1 (1=Air, 2=Car, 3 =Air+Car, 4 =Hotel, 5=Air+Hotel, 6=Hotel +Car, 7 =Air+Hotel+Car).
Line 3: We are creating the key-value pairs for those people who traveled by air, where the key is the destinations which are in 3rd column and value is 1. Since we need to count the popular cities, we are counting them by using the reduceByKey method. After counting the destinations, we are swapping the key-value pairs. We are using the sortByKey method to sort the data with keys where false stands for descending order. Once sorting is completed, we are considering top 20 cities that generate high airline revenues for travel.
Output:
(84,MIA), (68,SFO), (54,LAS), (42,LAX), (24,IAH), (23,DFW), (18,PHX), (17,BOS), (15,ORD), (13,NYC), (9,DCA), (8,WAS), (8,AUS), (7,DEN), (7,MEM), (7,JFK), (6,SYD), (6,PHL), (6,ATL), (5,RIC)
You can see the same in the below screen shot.
We hope this blog was useful. Keep visiting my blog https://hadoopbigdata123.blogspot.in/ for more updates on BigData and other technologies.
