nihadtp / covid19patientprediction

MAchie learning model developed to predict recovery of covid19 patient based on sex, age and state there living.

covid19PatientDeathPredictionProbability

This project uses Random Forest classifier module in spark ml library to predict probability of recovery of a covid19 patient based on there age, sex and state they are residing.

##Probability of recovery of different states from age 10 to 90

• Here x-axis is age and y-axis probability percentage.
• Here we can notice that probability decreases as age increases.
• Also probability of recovery is different for different states duu to difference in population, civic sense, hygiene etc
• By dividing whole dataset into 80-20 as training and test data. It is calculated to be accurate upto 93%
• Main data source are from covid19API in CSV format(Almost 50% data did not have sufficeint information)

Prequisites

1. Scala 2.11.12
2. Spark 2.4.6
3. SBT 1.0
4. Cassandra 4.0
5. cqlsh 5.0.1

Cassandra Set up

• Log into cassandra cqlsh and create Keyspace called covid19 and table called prediction

cqlsh> CREATE TABLE covid19.prediction(sex text, age float, state text, p_sex double, p_state double, p_state double,\ PRIMARY KEY (state_prop, country_prop, date, uuid));

Steps to build jar

1. First clone github project to a folder

2. Download csv data soircce to local covidAPI

3. In src/main/scala/covid19PredictionApp.scala file. Edit the following

val rawData = spark.read.format("csv")
      .option("header", "true")
      .load("/home/nihad/machine_learning/patient_data/*.csv")

Under load function function put in your correct path

2. Then build fat jar using sbt assembly plugin

sbt assembly

3. In target folder you will find covid19Prediction-assembly.jar file

4. Start your cassandra service

5. Go to spark folder

spark-submit ~/pathtoyourtargetfolder/covid19Prediction-assembly.jar

Here app starts and fetches data from the folder, converts them to Dataframe and trains using Random Forest classifier And predicted values are then written to cassandra table.