Predicting variable Time Series - Unknown Frequency

Question

I would like to create a model in order to predict the demand of a certain variable according to some historical data. I am working with Python and I am facing some problems.

1) My time series does not have a defined frequency. I have been working with R and time series in the past, and I used to specify a frequency for the observation. How should I deal with my case? (I can have 50 entries one day, 0 another day, and so on)

2) I would like to forecast the demand of a given "Function Title". Do you advise me to use any machine learning technique? (my dataset is about 26000 entries)

Here a preview of my dataframe

import pandas as pd
import numpy as np

df = pd.read_csv('dataset.csv')

#Replace column with english label
df.columns = ['Work Experience', 'Function Profile', 'Function Title',
              'Occupation', 'Education', 'Company Name', 'Clean Company Name',
              'Sector (Industry)', 'Location', 'Salary Indication', 'Province',
              'Date Found', 'ISCO', 'SBI-Code']
ts = df.set_index(dates)
ts.head()

Answer 1

The thing that is called a "time series" in normal statistical parlance always has regular values. There are techniques for filling in missing values (e.g. you want to analyze a daily temperature record with time series techniques, but there are some days the guy recording them was out sick). But this looks like an inherently stochastic problem: the time at which entry appears is inherently random.

If I understand your problem correctly, you can transform this into a time series by counting the number of entries per day (or week, or whatever). This count is then a time series, in the usual understanding of the term, and can be analyzed as such. And since what you want to predict is itself a number of entries in some future time period, this is probably the right approach.