A centralized dashboard for screening and downloading Stock/ETF/FX/Economic Data, as well as viewing Forex, Metal, Energy, and Crypto Calendars. Apart from being a one-stop data provider, Econ-Dashboard is also equipped with sentiment classifier trained on financial/economic sentiment. Time-series forecasting is armed with two LSTM models distinguished for different market capitalization constraints.
The sentiment classification model is available at my huggingface repository
The model is equipped with a pre-trained embedding layer(Universal-Sentence-Encoder), therefore, the model deployment requires a customized kerasmodel class.
# Define the custom layer
class USEEncoderLayer(tf.keras.layers.Layer):
def __init__(self, **kwargs):
super(USEEncoderLayer, self).__init__(**kwargs)
self.use_layer = hub.KerasLayer("https://tfhub.dev/google/universal-sentence-encoder/4",
input_shape=[], # check the important notes
dtype=tf.string,
trainable=False,
name="USE_encoder")
def call(self, inputs, **kwargs):
return self.use_layer(inputs)
# Register the custom layer
custom_objects = {"USEEncoderLayer": USEEncoderLayer, "KerasLayer": hub.KerasLayer}
# Load the model with custom layer
with tf.keras.utils.custom_object_scope(custom_objects):
best_use_model = tf.keras.models.load_model(r"WRITE_MODELS_PATH")
- Stock screener is equipped with yfinance API. You can select time frame and data interval, and download the data
- If you're not sure about which ticker/symbol format, head over to yahoo finance to check the ticker/symbol
- Data Provider is powered with various economic and financial data, including economic calendar, FX , Stock , Earnings, Commodities, Bonds, Crypto market data.
Sentiment Classification model is trained on Phrasebank's agreeall Financial sentiment dataset.
Forecasting model is a fine-tuned LSTM model that is trained on 7-days look-back period to forecast 1-day forward window value. The model inputs are; last available Volume and 7 consecutive daily close data. You have option to either upload your own data, or write the required data inputs on your own. The time-series feature is armed with two models that are specifically trained for both Large and Small Cap Stocks. You May Select The Model For Your Stock