At vero eos et accusamus et iusto odio dignissimos ducimus qui blanditiis praesentium voluptatum deleniti atque corrupti quos dolores et quas molestias excepturi sint occaecati cupiditate non provident, similique sunt in culpa qui officia deserunt mollitia animi, id est laborum et dolorum fuga.
Et harum quidem rerum facilis est et expedita distinctio. Nam libero tempore, cum soluta nobis est eligendi optio cumque nihil impedit quo minus id quod maxime placeat facere possimus, omnis voluptas assumenda est, omnis dolor repellendus.
Itaque earum rerum hic tenetur a sapiente delectus, ut aut reiciendis voluptatibus maiores alias consequatur aut perferendis doloribus asperiores repellat.
You can do a regression, but you will almost certainly have a very low correlation coefficient given the low number of data points, and the typical variation in the data from year to year. So don't put too much stock in the accuracy of the regression analysis.
Also, whether you are using regression or some other subjective or objective approach to weighting or trending the historical data, I would suggest forecasting working capital line-by-line with appropriate drivers for each line item. So accounts receivable and inventory are typically driven by sales. Accrued liabilities are driven by certain expense categories. I'd also recommend calculating the typical working capital ratios for the historical period such as inventory turnover, A/R days, A/P days etc. These help make more intuitive assumptions going forward and are more easily compared with peer companies. Technically, the most proper way to calculate the historical ratios uses average balances (e.g., sales / ((beginning inventory + ending inventory))/2), but this can cause problems when forecasting (see footnote below).
Capital expenditures are probably best forecast from discussions with management, or in the case of public companies, there is typically at least some disclosure of expectations about expansion through organic growth and acquisitions - whether from the company or research analysts.
Footnote: Let's take inventory turnover as an example. If you calculate inventory turnover using average balances of inventory, then your forecast of inventory would presumably be driven by a ratio interpreted the same way. So you would be using the forecasted inventory turnover to calculate the average balance of inventory in a given year. You then have to back into the ending inventory. This always creates a see-saw effect when forecasting, where one year is below the average, then the next year is above the average, etc. So then you have changes in working capital that swing dramatically from year to year. To avoid this problem, I tend to use a simpler version of these ratios which is calculated on ending balances only.