app.py

import streamlit as st
import datetime
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import plotly.express as px

import plotly.graph_objects as go
from plotly.subplots import make_subplots

st.set_page_config(layout="wide")


@st.cache_data
def create_datafram(start_date, num_prices, base_price, price_ratio):

    # Generate timestamps (assuming daily data)
    dates = pd.date_range(start=start_date, periods=num_prices)

    # Generate random prices 
    prices = [base_price]  # Initial value
    for _ in range(num_prices - 1):
        new_value = prices[-1] + (np.random.randn()* price_ratio)  # AR(1) process
        prices.append(new_value)

    # Create the DataFrame
    data = {
        'timestamp': dates,
        'price': prices
    }

    return pd.DataFrame(data)





st.header('Create Price Data', divider=True)

col1, col2, col3, col4 = st.columns(4)

with col1:
    date = st.date_input("Date Start", datetime.date(2019, 7, 6))
    
with col2:
    steps = st.number_input("Time steps (days)", value=365)
    
with col3:
    price = st.number_input("Base Price", value = 100)

with col4:
    ratio = st.number_input("Fluctuation factor [0 - 1]", value=0.9)
 
 
if 'data' not in st.session_state:
    st.session_state['data'] = create_datafram(date, steps, price, ratio)


if st.button('Create Time-series data'):
    st.session_state.data = create_datafram(date, steps, price, ratio)



st.header('Show Price Data', divider=True)


col5, col6 = st.columns((1,2))

with col5:
        
    st.dataframe(st.session_state.data, use_container_width=True)
            
with col6:
    
    fig = px.line(st.session_state.data, x="timestamp", y="price", title='Time Series Price Data')
    
    st.plotly_chart(fig, use_container_width=True)
    #st.line_chart(st.session_state.data, x="timestamp", y="price", use_container_width=True)
    



def cal_ipa(df_in: pd.DataFrame, window_size: int = 120, std_adjust : float = 2.0, ipa_col: str='IPA', adj_price: bool= False):

    if adj_price:
        
        df_in['price'] = df_in['price'] - (1 - df_in['price'].rolling(window=window_size).std())
        
    df_in[ipa_col] = abs((df_in['price'] - df_in['price'].rolling(window=window_size).mean()) / (df_in['price'].rolling(window=window_size).std()*std_adjust))
    
    return df_in

def cal_anomaly(df_in):
    
    return ...

    
st.header('Anomaly IPA Analysis', divider=True)

a_col1, a_col2, a_col3, a_col4 = st.columns(4)

with a_col1:
    window_quater = st.number_input("Quater window (days)", value=120)

with a_col2:
    window_month = st.number_input("Monthly window (days)", value=30)
    
with a_col3:
    std_adj = st.number_input("Standard Deviation Adjustment", value=2.5)
    
with a_col4:
    adj_price = st.toggle("Adjust Price")


b_col1, b_col2 = st.columns(2)

with b_col1:
    gamma = st.number_input("Gamma adjustment", value=0.8)   
    


if 'ipa_df' not in st.session_state:
    st.session_state['ipa_df'] = st.session_state.data.copy()
else:
    st.session_state.ipa_df = st.session_state.data.copy()
    
if not adj_price:
    st.session_state.ipa_df = st.session_state.data.copy()

st.session_state.ipa_df = cal_ipa(st.session_state.ipa_df, window_size=window_month, std_adjust=std_adj, ipa_col='IPA_m', adj_price=adj_price)
st.session_state.ipa_df = cal_ipa(st.session_state.ipa_df, window_size=window_quater, std_adjust=std_adj, ipa_col='IPA_q', adj_price=adj_price)

w_adj = np.random.uniform(low=0.9, high=1.0, size=steps)

st.session_state.ipa_df['IPA_q'] = st.session_state.ipa_df['IPA_q'] * w_adj
st.session_state.ipa_df['IPA_m'] = st.session_state.ipa_df['IPA_m'] * w_adj

st.session_state.ipa_df['ipa_adj'] = gamma* st.session_state.ipa_df['IPA_m'] + (1-gamma)*st.session_state.ipa_df['IPA_q']

st.session_state.ipa_df['Warning'] = 0
#df.loc[df['IPA'] < 0.5, 'Warning'] = 'Normal'
st.session_state.ipa_df.loc[st.session_state.ipa_df['ipa_adj'] >= 0.5, 'Warning'] = 1
st.session_state.ipa_df.loc[st.session_state.ipa_df['ipa_adj'] >= 1.0, 'Warning'] = 2


c_col1, c_col2 = st.columns((1,2))

with c_col1:
    st.dataframe(st.session_state.ipa_df)

with c_col2:
    # fig, ax1 = plt.subplots(figsize=(14, 5), dpi=160)
    # ax2 = ax1.twinx()

    # ax1.plot(st.session_state.ipa_df['price'])
    # #ax1.set_ylim(90, 150)
    # ax1.set_xlim(120, 365)

    # #ax2.plot(df['IPA'], c='red')
    # ax2.fill_between(st.session_state.ipa_df.index, st.session_state.ipa_df['ipa_adj'], color='blue', alpha=0.1)
    # ax2.fill_between(st.session_state.ipa_df.index, st.session_state.ipa_df['Warning'], color='orange', alpha=0.3)
    # ax2.set_xlim(120, 365)
    # ax2.set_ylim(0, 2)

    # st.pyplot(fig)
    
# =================================================================

    fig = go.Figure()

    fig.add_trace(go.Scatter(
        x=st.session_state.ipa_df['timestamp'], 
        y=st.session_state.ipa_df['price'],
        hoverinfo='x+y',
        mode='lines',
        line=dict(width=1.0, color='rgb(100, 100, 100)'),
        name='Price'
    ))

    fig.add_trace(
        go.Scatter(
            x=st.session_state.ipa_df['timestamp'], 
            y=st.session_state.ipa_df['ipa_adj'],
            hoverinfo='x+y',
            mode='lines',
            line=dict(width=1.0, color='rgb(0, 230, 230)'),
            stackgroup='one',
            name='IPA Value',
            yaxis='y2'
            )
        ,)

    fig.add_trace(
        go.Scatter(
            x=st.session_state.ipa_df['timestamp'], 
            y=st.session_state.ipa_df['Warning'],
            hoverinfo='x+y',
            mode='lines',
            line=dict(width=1.0, color='rgb(230, 172, 0)'),
            stackgroup='one',
            name='Anomaly Index',
            yaxis='y3'
            )
        ,)


    fig.update_layout(
        
        legend=dict(orientation="h"),
        
        yaxis=dict(
            title=dict(text="Price"),
            side="left",
            range=(min(st.session_state.ipa_df['price']*0.99), max(st.session_state.ipa_df['price']*1.01)),
        ),
        yaxis2=dict(
            title='',
            side="right",
            range=(0, 2),
            overlaying="y",
            showgrid=False,
        ),
        yaxis3=dict(
            title='Anomaly Index',
            side="right",
            range=(0, 2),
            overlaying="y",
            showgrid=False,
        ),
    )

    #fig.update_yaxes(title_text="<b>primary</b> yaxis title")
    #fig.update_yaxes(title_text="<b>secondary</b> yaxis title", secondary_y=True)

    fig.update_layout(
        title_text="IPA Anomaly detection chart",
    )
    st.plotly_chart(fig)