iyunbo / sentiment-detect Goto Github PK

This is WIP. Getting Started

Current infrastructure:

• TweetCollector serializes Tweets (without code generation) to Avro and sent to Kafka
• TweetAnalyzer picks up serialized Tweets and monitor tweets for unexpected volume in Spark
• Volume thresholds and detected alerts managed in HDFS

1. Get Twitter credentials and fill them in reference.conf.example and rename to reference.conf

2. Start Kafka (instructions) in single-node mode on localhost

3. Start TweetCollector

./gradlew collect 

This will start to read recent tweets, encode them to Avro and send to the Kafka cluster in binary format (Array[Byte]).

4. Start TweetAnalyzer

 ./gradlew analyze

This will run Spark streaming connected to the Kafka queue. In 5-second intervals the program reads tweets from Kafka, analyzes the tweet texts and print the 10 most tweeted company of the 400 S & P