import pandas as pd from nixtlats import TimeGPT import nltk from nltk.tokenize import word_tokenize import numpy as np from sklearn.preprocessing import StandardScaler import numpy as np from sklearn.metrics import mean_squared_error from math import sqrt import numpy as np from sklearn.metrics import mean_squared_error from math import sqrt from sklearn.metrics import classification_report,confusion_matrix timegpt = TimeGPT( # defaults to os.environ.get("TIMEGPT_TOKEN") token = 'my-token' ) data_path = "-----" df = pd.read_csv(data_path) df.shape def convert_to_datetime(date_string): try: date_object = datetime.fromisoformat(date_string.replace("Z", "+00:00")) return date_object except ValueError: return None df['_time'] = df['_time'].apply(convert_to_datetime) df['_time'] = df['_time'].dt.tz_localize(None) df.head() pm_df=df column_names=['Q_VFD1_Temperature','Description'] df=df[column_names] nltk.download('punkt') df['Tokenized_Description'] = df['Description'].apply(lambda x: word_tokenize(str(x)) if pd.notnull(x) else []) result = df[[column_names[0], 'Tokenized_Description']].values.tolist() all_tokens = set() _ = df['Tokenized_Description'].apply(lambda x: all_tokens.update(x) if x else all_tokens.add(None)) # List of all different tokens different_tokens = list(all_tokens) different_tokens tokens ={'NoBody1':2, 'E_STOPPED':3, ',':0, 'NoNose':1, 'None':0, 'NoBody2':2} def replace_with_numeric(tokens_dict, tokens_list): return [tokens_dict[token] if token in tokens_dict else None for token in tokens_list] df['Tokenized_Description1'] = df['Tokenized_Description'].apply(lambda x: replace_with_numeric(tokens, x)) def calculate_average(row): numeric_values = [value for value in row if value is not None] return sum(numeric_values) / len(numeric_values) if len(numeric_values) > 0 else 0 df['Average'] = df['Tokenized_Description1'].apply(calculate_average) features_df=df[[column_names[0],'Average']] features_df=features_df.rename(columns={"Average": "Description"}) features_df pm_df['Description']=features_df['Description'] """#splitting data""" train = pm_df[:int(0.8*(len(pm_df)))] valid = pm_df[int(0.8*(len(pm_df))):] """##creating training data""" train_2=train[column_names].round(0) train_2 """##Unique value count""" unique, counts = np.unique(train_2.Description.round(0), return_counts=True) dict(zip(unique, counts)) train_2.tail(20) """###FINAL TRAINING DATA train_new. This is the training data that we have used to train our model """ train_new=train_2.iloc[:95710] unique, counts = np.unique(train_new.Description.round(0), return_counts=True) dict(zip(unique, counts)) train_new.tail(20) """#SCALING DATA""" scalery = StandardScaler() # Fit and transform the data (for 'prediction') df_scaled = scalery.fit_transform(train_new.values) train_scaled = pd.DataFrame(df_scaled, index=train_new.index, columns=train_new.columns) train_scaled time=train['_time'].iloc[:95710] train_new_scaled=pd.concat([time,train_scaled],axis=1) train_new_scaled """#Creating new df which contains our final training data.""" new=train_new_scaled new=new.melt(id_vars=["_time"], var_name="unique_id", value_name="Value") new.rename(columns={'_time': 'ds' },inplace=True, errors='raise') train_new_scaled.tail(20) """#Predicting""" timegpt_fcst_pred_both_df = timegpt.forecast(model='timegpt-1-long-horizon',finetune_steps=90, df=new, h=200,freq='100ms', level=[90], time_col='ds', target_col='Value', ) ----> 2 timegpt_fcst_pred_both_df = timegpt.forecast(model='timegpt-1-long-horizon',finetune_steps=90, 3 df=new, h=200,freq='100ms', level=[90], 4 time_col='ds', target_col='Value', 5 ) 10 frames /usr/local/lib/python3.10/dist-packages/pandas/core/indexes/multi.py in from_arrays(cls, arrays, sortorder, names) 527 for i in range(1, len(arrays)): 528 if len(arrays[i]) != len(arrays[i - 1]): --> 529 raise ValueError("all arrays must be same length") 530 531 codes, levels = factorize_from_iterables(arrays) This code snippet is performing time series anomalay detection using the TimeGPT model from the Nixtla library on a dataset. Below is a breakdown of what the code does: Description of Data: - The

Description

column in the DataFrame contains the anomalies to be predicted. - The

Q_VFD1_Temperature

column is used as the feature for making predictions across the time series. - The

_time

column contains the time sequence dates. - The dataset Factory-Dataset is loaded into a DataFrame

df

. - The

_time

column is converted to datetime format and localized to remove timezone information. - The

Description

column is tokenized using NLTK's

word_tokenize

function. - Tokens in the

Description

column are mapped to numeric values using a predefined dictionary. - The average of these numeric values is computed for each row and added as a new column

'Average'

. - A new DataFrame

features_df

is created with the temperature column and the calculated average. - The data is split into training and validation sets. - The training data is further refined and scaled using

StandardScaler

. - The scaled data is reshaped and molded to create a DataFrame

new

suitable for forecasting using TimeGPT. - The

TimeGPT

model is used to forecast future values based on this prepared data. In the last line : timegpt_fcst_pred_both_df = timegpt.forecast(model='timegpt-1-long-horizon', finetune_steps=90, df=new, h=200, freq='100ms', level=[90], time_col='ds', target_col='Value') I am getting an error ValueError: all arrays must be of same length The error occurs when attempting to execute the forecast with TimeGPT, specifically at this line The traceback indicates that there is a mismatch in the lengths of arrays used during the forecast process. Can you please help me resolve the above error in the code ? The use case for which I would like to use TimeGPT is for the anomaly detection in Smart-Manufacturing domain?

@Harleen Kaur Bagga please edit this message to remove your token. This is sensitive information that should not be shared publicly.