INR-USD Timeseries Analysis

A time-series analysis for predicting INR value agains USD currency per date, using last 10 years Dataset with Neural Networks

Steps Performed

• Fetch Dataset
• Feature Extraction
• Data Analysis
• Preprocessing
• Building Model
• Training Model
• Result Analysis
• Hypertuning
• Saving Model

Model Build

Model: sequential Layer (type) Output Shape Param #
conv1d (Conv1D) (None, None, 60) 360
lstm (LSTM) (None, None, 60) 29040
lstm_1 (LSTM) (None, None, 60) 29040
dense (Dense) (None, None, 30) 1830
dense_1 (Dense) (None, None, 10) 310
dense_2 (Dense) (None, None, 1) 11
lambda (Lambda) (None, None, 1) 0
Total params: 60,591 Trainable params: 60,591 Non-trainable params: 0

---

Result

Choosing Learning Rate

Test 1

Test 2 - Hypertuned

Test 3

_____________ Accuracy Loss

For detailed info, check the notebook.