**TEAM MENTOR:**

- Subhabaha Pal (subhabaha.pal@manipalglobal.com)

**TEAM MEMBERS:**

- Suprio Dutta (17225760058@datascience.manipal.edu)
- Faraz Ahmad (17225760016@datascience.manipal.edu)
- Prankrishna Dolai (17225760039@datascience.manipal.edu)
- Sudarshan Mishra (17225760056@datascience.manipal.edu)
- Nitin Khandare (17225760036@datascience.manipal.edu)

**TEAM TOOLSET:**

- R
- Microsoft Excel

**BUSINESS UNDERSTANDING:**

The goal is to build a successful investing/trading model on the cryptocurrency markets. The data consists of time-series of various cryptocurrencies (in 5-minute steps) prices and 24 hour volumes.

**DATA UNDERSTANDING:**

- Data provided in
**csv**format - Exploring dataset’s structure
- Identifying discrepancies in the dataset
- Identifying the subsets of data that modelling would be based on

**DATA PREPARATION:**

**Raw Data:**

- Price_data à Data type: Time Series à Data Format: CSV
- Data have more than 1600 cryptocurrencies with transaction details
- Data provided between 17
^{th}JAN 2018 11:25 to 24^{th}MAR 2018 13:15

**Working:**

- Missing values in the dataset were identified and were imputed using “imputeTS” package.
- Converted the data to a time series by giving the frequency as 24 hours multiplied by 5 minutes.
- Variable reduction was done
- Here we have splited data base on traning and testing data set
- We have used here the neural network model
- By using forecast package we use the nnetar() function for forecasting time series
- This model makes the data stationary and then forecast it based on observation.

**Evaluation:**

- Divided the data into two parts namely training and test data set.
- Fed the model with the train data set that helped the model to learn something about the data.
- Predicted the values present in the train data set.
- Compared the Predicted values with the Observed values and calculated the RMSE( Root Mean Square Error).
- Lastly we have selected the model which has the least RMSE Value and higher Accuracy.

**Deployment:**

- Ran the algorithm on all the test data and created multiple CSV files with the required structure.
- Also did a manual check on the results with lowest confidence score to detect errors (and, as expected, found some which we manually corrected).